Master Art of Translation | The Food Composition, INFOODS & Nutrient Database Translation System — How Food Matching, Component Methods, Edible Portions, Recipes and Provenance Keep Their Meaning

How to translate food composition tables, INFOODS nutrient codes, food-matching records and recipe calculations without changing the data

How do you translate a food composition database, nutrition table, INFOODS nutrient tagname, or multilingual ingredient record accurately? The main difficulty is not selecting a fluent foreign-language word for protein, rice, fibre or energy. It is preserving exactly which food was sampled, which edible part was measured, what laboratory or calculation method produced a value, and what the number means per 100 grams, per serving, on a dry-matter basis or in a prepared dish. This guide explains food composition data translation as a reproducible evidence-preservation task rather than a vocabulary replacement exercise.

If you are searching for food composition table translation, FAO/INFOODS food component identifiers, EuroFIR LanguaL food description, EFSA FoodEx2 classification, USDA FoodData Central records, nutrient database harmonisation, food matching guidelines, cooking yield factors or nutrient retention factors, you need a method for separating language from scientific identity. Names, units, food states, methods, denominators, imputed data, detection limits and provenance must survive each stage of translation and database exchange. A readable translation can still be misleading when it silently merges distinct nutrient definitions or converts a food value to an incompatible basis.

This world-facing reference provides the hidden mechanism, a repeatable translation workflow, an extensive food-identity and nutrient-definition atlas, failure analyses, fictional record-by-record diagnostics, transfer workshops, quality checks and routes to authoritative documentation. It is designed for translators, nutrition researchers, food composition database compilers, localisation specialists, data stewards and editors working across languages. The worked records are expressly illustrative: invented values are not dietary advice, official reference numbers or laboratory findings. The goal is to make meaning traceable, not to substitute for a qualified nutrition scientist or a jurisdiction-specific labelling decision.

Series: EDKSG-TRANS-MASTER-WORLD-870. Research review: 1 November 2026. Scope: translation of food-composition and nutrient-database evidence, not general consumer food-safety messaging, individual dietary prescriptions or the ownership of vocabulary and English grammar lessons.

Root route: Master Art of Translation | The Complete System for Moving Meaning Between Languages. This child treats food composition records as its specialist object. The general translation system remains at the root, and protected learning-hub pages remain separate.

Quick navigation: the complete food-composition translation route

Search intent and scope
Hidden problems
Mechanism: how a food-composition assertion is built
Food-identity atlas
Component and measurement atlas
Repeatable production method
Failure modes and repairs
Worked fictional diagnostics
Transfer workshops
Working glossary
Questions and answers
Four-week implementation plan
Research and standards
Next learning routes

1. Search intent: decide what is being translated before editing the first cell

Some requests seek a human-readable bilingual food table for consumers. Others seek a machine-readable mapping that can combine nutrient values without altering the source record. A third group asks for a scientific interpretation of differences between databases. These jobs have different acceptance criteria. A public table may prioritise intelligible food names with retained scientific identifiers, whereas an integration service must protect every machine key even when the display label changes.

A food-composition table is an evidence system in miniature. A row identifies a food, sometimes with cultivar, origin, processing and preparation details. A column identifies a chemical or derived component. A value expresses an amount on a declared basis, obtained by measurement, calculation, literature compilation or estimation. A reference anchors the record in time and a methodological tradition. Translating only the row headings breaks the relationship between these parts.

The first question is whether the requested target is a faithful translation, a harmonised dataset, a dietary assessment input, an educational illustration, a product-label draft or a recipe calculation. Translation preserves source meaning; harmonisation may require explicit conversions; dietary interpretation adds assumptions about eating patterns. Those operations should be sequenced and documented rather than disguised under a single command labelled translate.

Multilingual search makes the task deceptively easy. A user searches for cooked beans and sees several matches whose local-language labels resemble one another. The candidates may differ in variety, salt addition, canning liquid, draining, edible fraction, fortification, moisture or analytical method. Choosing the first semantically similar name can introduce a systematic error that affects thousands of downstream estimates. Every search hit is a candidate, not a proof of equivalence.

When a nutrient name has no exact counterpart in the target language, the safest output may be a concise translated descriptive label plus the original component identifier and an explanatory method note. Refusing a false equivalence is not a translation failure. It is the moment at which the translation protects a scientific distinction that ordinary speech tends to blur.

A reader also needs to know whether the source is a national food composition database, a reference table, a branded-food label, a dietary survey compilation, a laboratory analysis or an experimental food study. These sources can all contain numbers for calcium or energy yet represent different populations and methodological decisions. The provenance class should appear before comparisons, averages or apparent rankings.

The output might have parallel language display fields without changing a stable key; the original and translation might share the same record identifier. Conversely, two independently sampled food records need separate identities even if their translated names are identical. A system that treats a translated title as a unique database identifier cannot distinguish linguistic variation from evidence variation.

Use a quality objective that can be tested: a competent reviewer should reconstruct the source food, component, measurement basis, methods, uncertainties, status codes and citation from the translated object. The reconstruction must be possible without trusting the translator’s prose. This criterion protects both human comprehension and machine interoperability.

2. The hidden problem: a nutrient number is meaningful only inside its record

A number such as 12 is not a free-standing scientific statement. It could mean 12 g protein per 100 g edible food, 12 mg iron per kilogram, 12 micrograms vitamin D per serving or a value below a quantification threshold. The unit, denominator, analytical method and status are parts of the assertion. If one of them is removed, the source may look easier to read while becoming less informative.

Food descriptions are context-rich. The same everyday word can refer to a species, a product category, a processing state, a prepared recipe, a commercial formulation or a traditional dish with varying ingredients. A translator who normalises all these cases to a familiar target-language word may accidentally substitute a different food. Better practice keeps a stable source label and records any target-language gloss as one view of the same identified entity.

Water is the great hidden variable. A 100 g portion of dried fruit is not comparable to 100 g of the corresponding fresh fruit unless moisture and the intended basis are explicitly considered. Cooking may increase or decrease mass while preserving, changing or redistributing nutrients. A more concentrated number per 100 g does not by itself prove that a nutrient was created or preserved.

The distinction between total carbohydrate, available carbohydrate, carbohydrate by difference, sugars and dietary fibre is often collapsed in everyday language. Different tables can calculate carbohydrate or energy using different conventions. Moving a value into a generic column with the label carbs can erase the calculation required to interpret it.

Protein is also partly a method problem. A value may be calculated from measured nitrogen with a declared conversion factor; it is not necessarily a direct assay of every intact protein molecule. If the factor changes, the translated numeric result cannot be represented as a verbatim value from the original source unless the recomputation is recorded separately.

A missing value is not a measured zero. A result below the laboratory’s limit of quantification is not always the same as a nondetection; neither necessarily means absence. A trace result, a blank, a not-applicable field and an inferred value need different data states. These distinctions matter especially when a downstream program interprets empty strings or zeroes without human judgment.

Analytical methods can disagree while both being legitimate for their intended targets. Fibre values, vitamin forms, fat extraction, mineral detection and energy calculations are not merely names. A table compiler must know which definition the source used before assigning a harmonised term. The translation should preserve uncertainty and method provenance rather than force agreement with a preferred terminology list.

Version control is another invisible source of meaning. A reference food can retain a familiar name while its database record, sampling basis, formulation or component calculation changes between releases. Historical values may remain legitimate for reproducing a published analysis. Automatically replacing them with today’s record silently changes the evidence rather than translating it.

Finally, classification vocabularies are not interchangeable. LanguaL expresses systematic facets of food descriptions. FoodEx2 provides hierarchies and facets for harmonised food description in a risk-assessment context. INFOODS tagnames identify components, not the same thing as a food category. USDA FoodData Central identifiers distinguish records and data types. A translator should retain these separate functions even when the interface shows them side by side.

The concealed ethical problem is overconfidence. Users may apply translated tables to public reports, education, procurement, research analyses or health communications. A translator cannot infer that a translated number is suitable for every such purpose. A good handover gives the recipient the meaning, conditions, exclusions and remaining verification work, rather than presenting polished prose as universal validation.

3. Mechanism: how food identity, components, values and evidence form one unit

1. The six-part assertion

Treat each nutrient claim as FOOD + COMPONENT + VALUE + UNIT/BASIS + METHOD/STATUS + SOURCE/VERSION. Translation is permitted to change the human-language display of the food and component when the new labels faithfully describe them. It must not silently modify the other pieces. The complete assertion is the minimum object that can be audited across languages.

Operational interpretation: during intake, mark every affected field as source-locked, translated-display, derived-with-method or unresolved. Source-locked data must reproduce the input exactly; translated display can change phrasing but cannot broaden or narrow what is identified; derived fields need an auditable calculation or mapping; unresolved material must be held for clarification. In review, test the assertion by asking what a second team would infer from the target-language record without seeing the original. If that inference differs, the conversion has altered evidence rather than language.

2. Food identity is a structured claim

Identify biological source, portion, processing, preparation, preservation, optional brand or recipe, and any meaningful population or sampling context. A name such as beans is a broad retrieval term, not the answer to whether the row represents dry seeds, cooked beans, a drained canned product or a seasoned dish.

3. Component identity is not a synonym list

The label protein may be serviceable for reading, but measurement and calculation qualifiers remain decisive. For component mapping, retain a source term, stable ID where available, target-language label, definition, analytical method and any factor. Mapping should be rejected or qualified when two source components have incompatible definitions.

4. The denominator travels with the value

The phrase per 100 g must specify the object whose 100 g are measured: edible raw portion, edible cooked portion, dry matter, as sold, or reconstituted product. If a value is per serving, record the serving weight and its source. No display-layer conversion should delete the original denominator.

5. Units can be precise while labels are vague

Micrograms, milligrams and grams are not interchangeable; international notation, decimal commas, multiplication symbols and superscript conventions can add ambiguity. The translation layer should show human-friendly units while the computational layer retains an unambiguous canonical representation.

6. Preparation is a transformation, not an adjective

Raw, boiled, steamed, fried, dried, fermented, fortified and drained imply processes that may change mass and composition. The translated descriptor should represent the action and end state, not an implied culinary judgment. The choice affects whether a food can be compared directly with another record.

7. A value has an epistemic status

Measured, calculated, estimated, borrowed, imputed, aggregated, imprecise, trace, below limit and not reported are different descriptions of knowledge. They answer how much confidence and comparability a reader should attach to a number. A display table without a provenance or status note often produces an appearance of certainty that the source never had.

8. A database release is part of the citation

A stable identifier is useful, but it is not always a timeless promise that the entire record has never changed. Preserve source, release, retrieval date and, where possible, a frozen record or checksum. If the content is updated, create a documented derived version instead of overwriting a historical translation.

9. A controlled vocabulary is not the food itself

A LanguaL descriptor or FoodEx2 code supports consistent retrieval and analysis; it does not guarantee that two food samples have identical composition. Classification equivalence is weaker than analytic equivalence. The translation needs to state when a code was assigned by a classifier, curator or machine and whether a human verified it.

10. Calculated values need a calculation ledger

For a recipe or energy calculation, retain source values, weights, yield and retention assumptions, equation, order of operations, rounding policy and final basis. Two translations that produce the same rounded number can still disagree about the ingredients or computation; reproducibility requires the ledger.

11. The source record is the authority for the translation

A translated term should never become evidence for changing a source nutrient value. When the data are internally inconsistent, record a discrepancy and seek the original analysis or database documentation. Translators may propose clarifications but should not silently repair scientific records as though they personally performed the assay.

12. Quality assurance has two tracks

Linguistic review checks accurate names, scope, modality, readability and audience. Data review checks identifiers, amounts, units, methods, reference states, formulae and lineage. Success on one track does not compensate for failure on the other. The two tracks should have separate sign-offs before release.

4. Food-identity atlas: translate the object before the nutrient value

The entries below are independent matching problems, not a universal lookup table. They teach how to protect food identity across languages and database taxonomies. Every food example names attributes to check, one plausible failure and an evidence-based route back to the source. The list deliberately spans raw ingredients, prepared dishes, commercial formulations and edible-part issues so that one routine cannot be overgeneralised.

01. Polished rice, raw versus boiled

Identity problem. The ordinary food name may omit water uptake and whether the grain was measured dry or cooked. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Require grain type, milling degree, raw/cooked state and a declared 100 g basis.

Misleading path. A dry-grain energy value presented under a cooked-rice label can inflate an apparent serving estimate. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Match the prepared-state row or show the dry state explicitly; preserve source ID and release.

Review and diagnostic. Ask whether water absorption or a changed 100 g denominator explains the apparent difference. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

02. Parboiled versus non-parboiled rice

Identity problem. Parboiling is a processing history, not merely a cooking instruction. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain the processing descriptor and grain form before matching similar names.

Misleading path. A target-language generic rice label can conceal a distinct micronutrient profile. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep parboiled as a separate searchable qualifier and check the original preparation.

Review and diagnostic. Never assume identical composition from shared cereal species. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

03. Dry chickpeas versus boiled chickpeas

Identity problem. The same legume name can conceal radically different water content and edible mass. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record botanical identity, dried/cooked form, soaking and added ingredients.

Misleading path. A translated table can mix dry-seed protein per 100 g with cooked-seed protein per 100 g. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep separate food identities and calculate only when yield assumptions are available.

Review and diagnostic. Trace the food state through every analysis and serving conversion. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

04. Canned beans drained versus in liquid

Identity problem. Drained solids differ from a record that includes the liquid. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Specify drained, rinsed, in brine or as sold; note added salt where reported.

Misleading path. A sodium comparison becomes misleading if one record includes packing liquid. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Match the edible fraction the consumer or survey actually records.

Review and diagnostic. Check whether a drained weight is present rather than inventing one. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

05. Wholegrain flour versus refined flour

Identity problem. Milling affects the plant parts represented, but language may reduce both to flour. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain cereal species, extraction rate or refinement status and enrichment.

Misleading path. Using a generic flour translation can distort fibre and mineral comparisons. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep wholegrain, refined and fortified descriptors separate.

Review and diagnostic. Review ash, fibre and fortification notes only as supporting evidence, not as a replacement for identity. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

06. Rolled oats versus instant flavoured porridge

Identity problem. A dry cereal ingredient is not the same as a sweetened ready-to-eat preparation. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record rolled/instant form, additions, reconstitution and water or milk.

Misleading path. Treating instant porridge as plain oats can erase added sugar and serving context. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Distinguish product formulation from base grain and preserve all source ingredients.

Review and diagnostic. Check whether a nutrient value refers to powder, prepared bowl or recipe. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

07. Wheat bread versus gluten-free bread

Identity problem. Functional food names can mask different ingredient matrices. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record grains or starch sources, fortification, fibre formulation and product form.

Misleading path. A translation that calls both simply bread can falsely suggest comparable nutrients. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep the product distinction and use descriptive terms when there is no exact local name.

Review and diagnostic. Do not infer allergen status or nutrient equivalence from a category alone. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

08. Fresh full-fat milk versus reconstituted powder

Identity problem. Concentration and preparation are essential to the serving basis. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Specify species, fat status, liquid/powder and reconstitution instructions.

Misleading path. Copying powder calcium per 100 g into the prepared beverage column is a denominator error. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Show whether values are as sold or as prepared, with any dilution assumptions.

Review and diagnostic. Verify whether added water or fortification explains differences. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

09. Fortified plant-based beverage

Identity problem. A familiar milk-like noun can hide added calcium, vitamin D and variable protein. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain source plant, fortification, sweetening, temperature treatment and product class.

Misleading path. Matching by target-language beverage name can replace a product-specific profile with a generic one. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Identify the actual formulation and source record rather than an assumed dairy analogue.

Review and diagnostic. Check that fortified values are not assigned to unfortified comparators. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

10. Strained yogurt versus regular yogurt

Identity problem. Straining changes water and solids, and the translated label may not encode it. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record milk source, fermentation, strained status, fat level and additives.

Misleading path. One product may seem unnaturally high in protein if its concentrated state disappears. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep strained or concentrated in the translated description and verify moisture basis.

Review and diagnostic. Do not use marketing words as analytical method descriptions. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

11. Hard cheese versus fresh cheese

Identity problem. Different cheeses are not a single nutrient reference item. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain ripening, moisture, milk species, salting and formulation where known.

Misleading path. Replacing cheese variety with a broad generic name weakens sodium and calcium comparisons. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Prefer the source’s specific cheese identity and add an explanatory gloss.

Review and diagnostic. Avoid claiming a protected origin or method that the record does not establish. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

12. Raw meat cut versus roasted lean meat

Identity problem. Cut, trimming and cooking alter the edible fraction and resulting mass. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Preserve species, anatomical cut, fat trimming, bone status and cooking treatment.

Misleading path. A translated label that drops cooked or lean changes the referent before any calculation. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Map to a record with matching cut and preparation, otherwise mark approximate.

Review and diagnostic. Check whether fat was removed before or after heating. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

13. Chicken skin on versus skin removed

Identity problem. Skin is an edible-part distinction with strong energy implications. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Keep meat-only, meat-plus-skin, bone removed and cooking state separate.

Misleading path. A shortened translated name can hide the portion contributing fat. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Carry the skin qualifier into tables, charts and recipe ingredients.

Review and diagnostic. Check portion-weight files because bone and skin can alter the served edible mass. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

14. Fish canned in oil versus water

Identity problem. Packing medium may be eaten, drained or excluded. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Preserve species, packing medium, drained status and brine or salt.

Misleading path. Interchanging canned-in-oil and canned-in-water records creates false fat differences. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Identify the exact product state and state how packing liquid was treated.

Review and diagnostic. Never assume a drained nutrient value means the can’s full mass. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

15. Smoked fish versus fresh fish

Identity problem. Smoking combines processing and potential salting and moisture changes. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Specify species, smoking method if known, salt condition and product form.

Misleading path. A generic fish translation may remove a high-impact processing descriptor. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep the source preservation method and reference context.

Review and diagnostic. Separate the composition issue from any unrelated food-safety claim. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

16. Bivalve flesh versus whole shellfish

Identity problem. Shells are not normally part of edible nutrient values. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. State edible flesh, shell-on purchase weight, cooking and liquid inclusion.

Misleading path. Using shell-on weights as edible grams depresses per-serving estimates. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Attach edible-portion conversion only when an authoritative factor is given.

Review and diagnostic. Check whether the sample represents meat, liquor or combined contents. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

17. Shrimp raw peeled versus shell-on

Identity problem. Peeling and head removal affect yield and reference mass. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record shell, head, tail, deveining and cooking state as available.

Misleading path. A target-language shrimp label cannot determine how much of a purchased weight is edible. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep purchase and edible weights in different fields.

Review and diagnostic. Use specified edible fraction, never a universal percentage invented by translation. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

18. Egg whole versus yolk versus white

Identity problem. The biological object is divided into materially different food fractions. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain whole egg, yolk, white, raw or cooked and any added fat.

Misleading path. A translator who collapses parts to egg destroys comparison of fat and protein. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Preserve fraction and preparation as primary identity terms.

Review and diagnostic. Check serving weights and whether the record excludes shell. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

19. Mushrooms grown with ultraviolet exposure

Identity problem. Vitamin D can depend on light exposure and preparation. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record species, cultivation/exposure method, fresh/dried form and date.

Misleading path. A general mushroom category may imply a profile from a very different sample. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep exposure as provenance rather than assuming all mushrooms are identical.

Review and diagnostic. Verify which vitamin D form is reported before comparing it. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

20. Spinach boiled and drained

Identity problem. Cooking water and leaf moisture affect reported values. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain raw, boiled, drained, salt-added or unseasoned state.

Misleading path. A translation that substitutes raw spinach can create a false vitamin change. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Track the prepared state and whether cooking liquid was discarded.

Review and diagnostic. Distinguish a per-100-g concentration shift from total nutrient retention. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

21. Carrots steamed versus fried

Identity problem. Heat method and fat addition may change more than flavour. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Specify preparation, added oil, peeling and cooked weight basis.

Misleading path. A generic cooked-carrot label can conceal ingredient additions. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Preserve the cooking method and distinguish a composite recipe from a single food.

Review and diagnostic. Check whether oil is counted in the food record or separately. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

22. Tomatoes fresh versus sun-dried

Identity problem. Dehydration shifts nutrients per unit mass and may involve oil or salt. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record fresh/dried, packing medium and rehydration where relevant.

Misleading path. A dry-tomato concentration may be misread as a biological increase. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep preparation and moisture status paired with each numeric column.

Review and diagnostic. Ask whether the value is for dried product or after reconstitution. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

23. Apple with peel versus peeled

Identity problem. Peel changes the edible part while the fruit name stays constant. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Preserve variety if documented, peel status, raw/cooked and sampling basis.

Misleading path. Matching an unpeeled apple to peeled values changes fibre interpretation. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Use explicit peel descriptors rather than a presumed standard fruit.

Review and diagnostic. Check whether an apple portion weight includes core or refuse. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

24. Dried apricots versus fresh apricots

Identity problem. Concentration and preservative treatment are both possible distinctions. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record dehydration, added sulfur compounds if listed and edible fraction.

Misleading path. Literal fruit-name translation can collapse fresh and dried samples. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep dried state and special treatment visible in all language variants.

Review and diagnostic. Do not infer treatment solely from product colour. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

25. Roasted salted nuts versus raw unsalted nuts

Identity problem. Roasting changes mass and added salt changes sodium. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Keep nut species, roasted/raw, salt addition, oil treatment and shell status.

Misleading path. Generic nuts creates a misleading sodium or energy comparison. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Show state and additions as part of the name and metadata.

Review and diagnostic. Check if values apply to kernels rather than shell-on purchases. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

26. Nut butter with sugar and oil additions

Identity problem. A processed spread may not match ground nuts alone. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record ingredients, sugar or salt, stabilisers and brand formulation.

Misleading path. Replacing spread with pure nut paste obscures the component sources. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Separate product-specific formulation from plain ingredient records.

Review and diagnostic. If formulation is unknown, avoid invented proportions. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

27. Cold-pressed oil versus refined oil

Identity problem. Processing, oil source and micronutrient profiles may differ. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain botanical source, refining state and use of blends.

Misleading path. A target label such as vegetable oil loses identifying information. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep each oil type distinct and mark blended products as blends.

Review and diagnostic. Check that fatty-acid profiles and composition are not copied between oils. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

28. Sweetened breakfast cereal with enrichment

Identity problem. A cereal brand may change formulation over time. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record brand, product name, ingredient formulation, fortification and data date.

Misleading path. An old branded label translated as a current product may misinform a comparison. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep label date and version; use nutrient data from the correct release.

Review and diagnostic. Never substitute marketing descriptions for component definitions. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

29. Dried seaweed versus hydrated seaweed

Identity problem. Water uptake and species matter strongly in seaweed records. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Specify species where known, dried/hydrated form and preparation.

Misleading path. Moving a dried value into a hydrated serving creates concentration errors. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Make the dry or hydrated basis visible before converting masses.

Review and diagnostic. Flag iodine variability and do not infer medical suitability from the translation. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

30. Broth concentrated versus diluted

Identity problem. A concentrate’s per-100-g values are not those of its prepared soup. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record dilution ratio, broth ingredients, salt and prepared state.

Misleading path. Serving values can be overstated if concentrate is treated as the consumed liquid. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Preserve concentrate and preparation instructions as separate data.

Review and diagnostic. Check actual water addition and final mass before any computed value. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

31. Salad with dressing versus plain leaves

Identity problem. The composite dish may include oil, nuts, cheese and sauces. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record all included components and portion definition.

Misleading path. A brief salad translation can replace a recipe with a vegetable-only profile. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep a recipe identifier or mark the match as non-equivalent.

Review and diagnostic. Require ingredients and weights if an aggregate calculation is requested. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

32. Fermented vegetables in brine

Identity problem. Fermentation, salting and draining are distinct attributes. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain vegetable source, brine state, fermentation and any seasoning.

Misleading path. Generic pickled vegetable labels blur vinegar pickle and microbial fermentation. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Name the process only when supported by the source record.

Review and diagnostic. Check whether sodium refers to solids or solids with liquid. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

33. Breaded fried fish portion

Identity problem. A composite processed food is not simply fish cooked in oil. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain batter or breading, fat uptake, species and edible portion.

Misleading path. Simplifying to fried fish can erase starch and oil contribution. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Use the complete prepared-food descriptor and source recipe.

Review and diagnostic. Differentiate recipe-derived numbers from direct food analysis. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

34. Infant formula powder versus prepared feed

Identity problem. The product state, dilution and intended population are critical. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record as-sold powder, reconstituted concentration and manufacturer’s stated preparation.

Misleading path. Copying dry powder values to liquid-feed units creates extreme errors. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Do not adapt feeding instructions or ingredient amounts without specialist review.

Review and diagnostic. Preserve exact product version and any instructions as source quotations. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

35. Tea leaves versus brewed tea

Identity problem. Beverage infusion and dry leaf have different measurement objects. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain infusion time, concentration, water volume and preparation.

Misleading path. A dry-leaf nutrient profile is not a cup-of-tea profile. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Match to the prepared infusion or declare lack of a comparable value.

Review and diagnostic. No assumption about extraction fraction belongs in a translation alone. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

36. Coffee grounds versus ready-to-drink coffee

Identity problem. Extraction removes some components into the beverage. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record ground/soluble/brewed form, added milk or sugar and serving basis.

Misleading path. Generic coffee translations can confuse concentrated powder and diluted drink. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep product and preparation states separate.

Review and diagnostic. Check caffeine units and the stated volume or mass denominator. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

37. Fortified flour versus unenriched flour

Identity problem. Added micronutrients require a formulation or regulatory context. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record enrichment/fortification and the relevant source period.

Misleading path. The same grain name can carry very different iron and folate values. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Preserve fortification status explicitly in each translated food label.

Review and diagnostic. Do not assume the same requirements across markets or time. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

38. Plant protein isolate versus whole seed

Identity problem. An isolate is not nutritionally identical to the crop it derives from. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record fractionation, protein concentration, moisture and additives.

Misleading path. Calling an isolate the whole seed changes the food’s identity. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep ingredient processing terms and product source separate.

Review and diagnostic. Check protein determination factor where the method is available. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

39. Mixed grain dish with a named recipe

Identity problem. A culinary name may cover many household variants. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain actual ingredients, preparation and recipe version when present.

Misleading path. Matching by a common dish name can produce systematic misclassification. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Attach a descriptive recipe gloss and do not infer fixed ingredients.

Review and diagnostic. Treat recipe-to-database mapping as approximate unless evidence supports equivalence. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

40. Frozen berries with added syrup

Identity problem. Freezing alone does not imply sweetness or added water. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record plain frozen, sweetened, syrup-packed or thawed state.

Misleading path. Target labels that omit syrup can hide substantial sugar contribution. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Preserve packing and thawing conditions.

Review and diagnostic. Check whether values include the packing liquid. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

41. Mineral water and fortified beverage

Identity problem. Natural mineral content and added minerals may be described similarly. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record water source or beverage formulation, carbonation and fortification.

Misleading path. A general water label can imply identical mineral values across products. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Retain product type and data source without assigning health benefits.

Review and diagnostic. Check whether values are given per litre, 100 mL or 100 g. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

42. Dried spice used in small quantities

Identity problem. High per-100-g concentrations do not describe the usual amount eaten. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record dried/ground form and intended portion if applicable.

Misleading path. A translated chart may mislead by comparing 100 g of spice with 100 g of a staple. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep the analytical reference basis and avoid unsupported portion interpretation.

Review and diagnostic. Separate food composition from consumption frequency. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

43. Raw cassava versus processed cassava product

Identity problem. Processing and cultivar can be important to the referent. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record raw/processed state, edible portion and product matrix.

Misleading path. An unqualified cassava translation can conflate flour, cooked root and raw material. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep precise identity without drawing safety conclusions from an isolated nutrient table.

Review and diagnostic. Refer any processing safety question to specialist guidance. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

44. Coconut flesh, milk and water

Identity problem. The common coconut name covers anatomically and compositionally distinct foods. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Specify flesh, expressed milk/cream or liquid water; note dilution.

Misleading path. A coconut entry cannot be matched on name alone. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Preserve product fraction and processing separately.

Review and diagnostic. Check moisture and fat because they signal different product types but do not prove identity. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

45. Sesame seed versus sesame paste

Identity problem. Grinding and added ingredients can alter usable comparisons. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record whole seeds, hull status, paste form and salt.

Misleading path. A simple seed translation may hide dehulling and preparation. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Keep paste and seed as different food states.

Review and diagnostic. Do not manufacture a density conversion between tablespoons and grams. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

46. Cooked lentil soup versus plain boiled lentils

Identity problem. Added water and vegetables produce a composite recipe. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Record liquid volume, ingredients, cooked yield and seasoning.

Misleading path. Lentils in soup are not numerically interchangeable with cooked lentils per 100 g. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Retain recipe identity, then compute ingredient contributions with documented methods.

Review and diagnostic. Trace each nutrient through recipe yield and retention assumptions. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

47. Ready-to-eat frozen meal

Identity problem. Commercial preparation may include sauce and multiple food groups. A food label can work as a human search term while failing as a scientific join key. The source must be treated as a qualified record with a specific edible object and state, rather than a dictionary entry that promises the same referent in every language. Retain product identity, instructions and as-sold versus heated state.

Misleading path. Replacing the meal with a broad dish label loses source formulation. The result can appear fluent, especially when a chart hides source identifiers and preparation details. It fails because the translated name allows an interpretation the input evidence did not license. Use a versioned product record and explicit preparation state.

Review and diagnostic. Do not infer current manufacturer composition from an old label. Compare candidate records on biological or recipe identity, raw versus prepared state, edible portion, data type, unit denominator and date. If one decisive attribute is missing, mark the matching claim unresolved and keep the literal source label alongside the proposed target display. This restraint is preferable to a confident but false nutrient equivalence.

5. Component and measurement atlas: name the chemistry and the method

This atlas separates ordinary translation labels from data definitions. Official INFOODS tagnames and source database codes should be copied from their current registry or record, not guessed from the English name. Examples here identify semantic categories and decision rules; they do not constitute an alternative controlled vocabulary.

01. Energy in kilocalories

Definition and boundary. Energy is often a calculated quantity, and a kcal value may depend on component definitions and conversion factors. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not translate energy as an independently measured calorimetric result unless the source explicitly says so. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain exact kcal/100 g versus kcal/serving basis and calculation metadata. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

02. Energy in kilojoules

Definition and boundary. kJ and kcal are distinct units linked by a defined conversion; displayed values may also reflect rounding or different calculation conventions. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not assume any published kJ and kcal pair must match a simple conversion exactly after independent source rounding. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain original values and note whether recomputation is a derived field. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

03. Water or moisture

Definition and boundary. Water mass fraction depends on preparation, drying and measurement methods. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not treat water percentage on a dry matter basis as edible-food moisture by default. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Check whether values use 100 g as eaten or an explicitly dried preparation. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

04. Protein from nitrogen

Definition and boundary. Protein estimates can use measured nitrogen multiplied by a food-dependent factor. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Never silently replace the source factor with 6.25 or with another customary number. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve method, conversion factor and numerical basis when available. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

05. Total fat

Definition and boundary. Extraction techniques and the definition of total lipids affect comparability. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Avoid translating different extraction results as one fully interchangeable measure. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep source component definition and analytical method tied to the value. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

06. Carbohydrate by difference

Definition and boundary. This derived figure can reflect water, protein, fat, ash and other included terms. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not equate it automatically with analytically measured available carbohydrate. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep the equation or source code when a comparison requires the distinction. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

07. Available carbohydrate

Definition and boundary. The concept may exclude fractions that total carbohydrate includes and may be calculated in different ways. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not collapse available carbohydrate, net carbs and total carbs without an exact definition. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Check fibre handling and the database’s component term. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

08. Total sugars

Definition and boundary. Total sugars can mean an analytical sum of particular mono- and disaccharides. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer added sugars from total sugars. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain method or constituent definition and separate sugar claims. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

09. Added sugars

Definition and boundary. A formulation-based category is not automatically recoverable from one chemical measurement. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer added-sugar quantity from a total-sugar number or from a translated product name. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Require source formulation or a documented assessment method. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

10. Dietary fibre

Definition and boundary. Different laboratory methods and component definitions can generate different fibre values. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not equate a generic fibre label across tables when tagnames or methods differ. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep the source method and component identifier. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

11. Total starch

Definition and boundary. Measured starch is distinct from total carbohydrate and from resistant starch classifications. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not replace a source starch result with carbohydrate by difference. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve whether the figure is total starch or a specific fraction. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

12. Resistant starch

Definition and boundary. Fraction depends on method and processing history. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer resistant starch from a food’s familiar name or cooling history alone. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Require analytic basis and processing documentation. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

13. Ash

Definition and boundary. Ash is a residue measure used in food analysis, not a nutritional mineral amount. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not translate the word as though it were identical to total mineral intake. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep method and dry/wet basis. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

14. Sodium

Definition and boundary. Mass of sodium is not the same thing as mass of salt. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not replace mg sodium with mg salt; converting salt equivalents requires a declared convention. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain source units and the wording of any derived salt equivalent. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

15. Potassium

Definition and boundary. Mineral concentration depends on food and preparation, and loss into cooking liquid can matter. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not compare raw and boiled records without their preparation states. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep mg/100 g and data source separate. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

16. Calcium

Definition and boundary. Fortification, food fraction and analytical source can change calcium values. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not present a fortified beverage’s calcium as an intrinsic property of every version. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep food formulation and analytical provenance. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

17. Iron

Definition and boundary. Total chemical iron is not a clinical measure of absorbed iron. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not translate food iron content as equivalent to bioavailable or absorbed iron. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Reserve physiological interpretation for a separate specialist assessment. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

18. Zinc

Definition and boundary. Values require source food, unit and sampling information. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer intake adequacy merely from a concentration listed in a table. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep the component and study context. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

19. Iodine

Definition and boundary. Marine food composition can vary markedly with species and processing. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not use a single generic seaweed value as though it described all products. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve species, dry/wet basis and method details. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

20. Selenium

Definition and boundary. A trace element’s concentration may vary with growing conditions and source. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not suppress geographic or sample provenance when it explains meaningful variance. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Maintain scientific context while keeping the article globally framed. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

21. Vitamin A retinol

Definition and boundary. Preformed retinol is not the same component as total vitamin A activity. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not merge retinol and carotenoid equivalents into one unnamed amount. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve the source basis and biochemical form. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

22. Vitamin A activity equivalents

Definition and boundary. Retinol equivalents and retinol activity equivalents can use different conversion schemes. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not equate numerical values merely because both translated labels contain vitamin A. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Identify the equivalence basis and computation. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

23. Beta-carotene

Definition and boundary. A provitamin A carotenoid has its own measurement and unit. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not treat beta-carotene mass as identical to retinol mass or activity. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep the analyte identity and any conversion separate. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

24. Vitamin D

Definition and boundary. D2, D3 and other reported forms or totals need explicit definitions. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not compare vitamin D entries until the included forms and units are known. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Document food exposure and fortification when relevant. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

25. Vitamin E

Definition and boundary. Alpha-tocopherol and total tocopherol measures are not equivalent. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not reduce multiple vitamer measurements to an unnamed vitamin E number. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain biochemical form and component definition. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

26. Thiamin

Definition and boundary. Vitamin B1 may be presented in milligrams, often at low concentrations. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not lose the decimal position in locale conversion. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Protect mg units and numeric precision. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

27. Riboflavin

Definition and boundary. Vitamin B2 reporting may involve low values and differing food preparation. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not assume a blank field means the vitamin was absent. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Check status and measurement method. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

28. Niacin

Definition and boundary. Preformed niacin and niacin equivalents can represent different calculations. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not use the equivalent amount as if it were simply measured niacin. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve conversion assumptions and component title. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

29. Folate

Definition and boundary. Food folate, folic acid and dietary folate equivalents do not have identical meanings. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not translate a fortified-food folic-acid value as naturally occurring folate. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain biochemical basis and equivalent-unit conventions. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

30. Vitamin B12

Definition and boundary. Different food types and fortification can affect interpretation. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer biological availability or nutritional outcome from concentration alone. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep µg/100 g and record preparation. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

31. Vitamin C

Definition and boundary. Processing and storage can change ascorbic-acid-related values. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not compare values without noting preparation and dataset date. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve analytically reported component and method. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

32. Total saturated fatty acids

Definition and boundary. A sum depends on which fatty acids are included and the source’s conventions. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not treat saturation categories as synonymous with total fat. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep composition fraction and denominator. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

33. Total trans fatty acids

Definition and boundary. The term can cover particular fatty-acid isomers and reporting conventions. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer origin or product compliance from the total concentration alone. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Document constituent definition and analytical limits. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

34. Omega-3 fatty acids

Definition and boundary. ALA, EPA and DHA are not chemically interchangeable. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not relabel a sum as though it were EPA+DHA when ALA is included. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep every individual or aggregate definition. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

35. Cholesterol

Definition and boundary. This sterol may be reported as mg/100 g edible portion. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not assume all plant-food database blanks equal a laboratory-confirmed zero. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Distinguish chemical absence assumptions from measured evidence. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

36. Alcohol

Definition and boundary. Ethanol may contribute energy and can vary by product or preparation. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not silently omit its contribution from a declared energy calculation. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve the product and calculation context. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

37. Organic acids

Definition and boundary. Some energy conventions account for organic acids separately. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not re-compute energy with one factor set while leaving the source label unchanged. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Record the chosen standard and its factor ledger. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

38. Polyols

Definition and boundary. Sugar alcohols are a distinct carbohydrate-related group. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not automatically assign ordinary carbohydrate energy factors to every polyol. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve specific composition and source convention. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

39. Edible portion coefficient

Definition and boundary. A proportion links purchased mass with the fraction intended for consumption. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not report an edible-fraction coefficient as a nutrient concentration. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep the fraction, food condition and method explicit. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

40. Recipe yield factor

Definition and boundary. Yield relates starting ingredient weight to final dish weight. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not confuse it with nutrient retention or with a percentage of recipe ingredients. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Maintain full starting and final mass records. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

41. Nutrient retention factor

Definition and boundary. Retention estimates nutrient remaining under an identified processing procedure. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not use a retention factor as a multiplier on concentration without accounting for recipe mass. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep factor definition, heat method and source. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

42. Analytical value

Definition and boundary. An observed lab value is tied to samples and methods. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not reclassify it as a universal nutrient requirement or a guaranteed product property. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep sample counts and statistical summaries if provided. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

43. Imputed value

Definition and boundary. A gap filled using a documented inference is not equivalent to direct measurement. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not hide the estimation lineage behind a clean numeric cell. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Carry imputation flag and source reasoning. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

44. Limit of quantification

Definition and boundary. LOQ marks a measurement capability boundary for quantifying an analyte. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not round ‘<0.2’ to zero or translate it as undetectable without verifying the source terminology. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep comparator sign, unit and analytic context. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

45. Trace value

Definition and boundary. A trace may signify detected but not quantified or a database-specific status. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not convert the word trace into a fixed number. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain the source’s data status and definition. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

46. Missing or not reported

Definition and boundary. No value supplied is a knowledge gap, not necessarily a chemical zero. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not fill blank cells with zero for display convenience. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Distinguish missing, unknown, not applicable and below limit. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

47. Sample variation

Definition and boundary. A nutrient mean may aggregate multiple samples with an associated distribution. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not suppress sample size, variance or range when they are important to interpretation. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Keep summary statistic type and sampling period. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

48. Data release and version

Definition and boundary. Record identifiers can outlive changes to source content. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not replace a previously used dataset with today’s values while claiming a verbatim translation. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain source URL, ID, revision and access date. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

49. Food component tagname

Definition and boundary. An INFOODS tagname is an identifier for a defined food component, not a general-language replacement. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not translate the machine identifier itself into a natural-language string. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Translate the display label while maintaining the original code. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

50. Food classification facet

Definition and boundary. A facet supplies a structured qualifier for food description. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not infer that a facet is a measured analyte. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve hierarchy, assignment method and meaning of the code. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

51. Branded product label

Definition and boundary. Manufacturer-provided label data and analytical reference data are not identical data types. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not treat each branded label figure as an independent new laboratory assay. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Record product formulation and label date. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

52. Density and volume conversion

Definition and boundary. A volume-based portion requires a product-specific mass or density assumption. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not treat 100 mL as exactly 100 g for every liquid. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Preserve the measured or documented density when making conversions. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

53. Decimal and grouping separator

Definition and boundary. A comma can be a decimal symbol in one language and a thousands separator in another. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not change numerical magnitude when localising display punctuation. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Use machine-safe numerical storage distinct from locale formatting. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

54. Micronutrient precision

Definition and boundary. A small microgram value can be materially altered by inappropriate rounding. An acceptable translated display label must tell a reader which food component or property is meant, while leaving any stable identifier intact. This is not just terminological polish: a scientific database join assumes that matched fields denote sufficiently comparable things.

Common error. Do not introduce significant digits that the source did not report. The error is difficult to see in a multilingual table because the numeric value remains plausible after the wording changes. When a component term is broad or ambiguous, show a concise method qualifier rather than forcing an apparent one-to-one translation.

Validation gate. Retain raw numeric precision and uncertainty notes. Preserve source value, original units, reference basis and provenance alongside the target label. If no explicit field or source note resolves the methodological distinction, report the mapping as provisional and exclude it from automated arithmetic or pooled comparisons until a subject expert reviews it.

6. Repeatable production method: twenty-six gates from source to verified translation

Gate 01: Define the deliverable

Purpose. State whether the output is a translated display, cross-database mapping, merged data product, recipe calculation or editorial explainer. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Require a signed purpose and allowed transformations. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Reject vague requests to make all foods equivalent across languages. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 02: Freeze the source

Purpose. Record the database, release, retrieval timestamp, specific record IDs and source file hashes where available. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Create a read-only snapshot and a working copy. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A moving source cannot support an exact reproducible translation. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 03: Identify audiences

Purpose. Differentiate public readers, nutrition scientists, developers, curators and legal labelling reviewers. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Agree terminology depth, exceptions and retained identifiers. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A short user label must not be mistaken for a fully harmonised analytic field. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 04: Inventory identifiers

Purpose. List food IDs, component IDs, method codes, language fields, version IDs and references. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Tag stable keys as non-translatable and verify round-tripping. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. The same visible phrase may name more than one record. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 05: Extract food identity

Purpose. Record species or food source, edible fraction, preparation, processing, preservation, recipe or brand. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Use a structured intake sheet rather than a prose-only note. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. An unqualified food name is not a reliable join key. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 06: Extract component definitions

Purpose. For every numeric column capture original component term, code, definition and method qualifiers. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Make a component crosswalk with approved, provisional and incompatible states. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A label match is not enough to establish analytic comparability. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 07: Classify value provenance

Purpose. Mark measured, calculated, estimated, imputed, literature-derived, manufacturer label or missing. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Keep status in an auditable column. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Values from unlike origins must not be merged without disclosure. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 08: Protect numeric strings

Purpose. Parse sign, decimal separator, magnitude, exponent, uncertainty and comparator separately. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Test reserialization before applying locale display rules. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A correct translation of the column heading cannot rescue a tenfold number error. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 09: Protect units and basis

Purpose. Distinguish g, mg, µg, kcal, kJ, per 100 g, per serving, dry matter and as consumed. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Require an explicit basis on every quantitative assertion. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Unit normalization must be logged as a derived change, not disguised as translation. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 10: Audit missingness

Purpose. Catalogue blanks, trace marks, limits of detection and limits of quantification using the source’s dictionary. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Prevent default-to-zero data transformations. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Data status can be more informative than a rounded number. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 11: Map food descriptions

Purpose. Use source labels, standard descriptors, LanguaL facets or FoodEx2 codes where appropriate. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Retain unmatched attributes and record match confidence. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Codes support classification, not an automatic claim of identical composition. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 12: Map language terms

Purpose. Translate display labels with a controlled glossary, examples and context notes. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Review false friends, culinary ambiguity and regional synonyms. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A readable term should never override the structured food definition. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 13: Validate recipe ingredients

Purpose. For composite dishes freeze ingredient identities, quantities and preparation states. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Distinguish measured dish results from calculated recipes. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Replacing an ingredient by a similar named product changes the recipe evidence. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 14: Verify edible yields

Purpose. Track purchased quantity, peel, bone, shell, trimming and edible weight. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Calculate only when the fraction’s source and applicability are documented. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A package weight is not always an edible-food weight. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 15: Verify cooking yields

Purpose. Record starting and final recipe masses and distinguish water changes from nutrient changes. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Check that every per-100-g result refers to the final mass if labelled cooked. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A yield factor is not the same as nutrient retention. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 16: Verify nutrient retention

Purpose. Retain factor table, cooking method, ingredient component and any assumptions. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Check each component independently instead of applying one universal percentage. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Retention can differ by nutrient and processing method. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 17: Evaluate energy conventions

Purpose. Determine whether energy uses specified factors, carbohydrate definitions and fibre handling. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Do not silently recompute a source-reported energy value. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Two legitimate calculations may differ because their definitions differ. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 18: Review micronutrient forms

Purpose. Check retinol, vitamin A equivalence, folate conventions, vitamin D forms, niacin equivalents and others. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Build form-specific mappings, not just a single vitamin-name column. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Different chemical or activity measures may share a familiar label. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 19: Test representative samples

Purpose. Choose foods with moisture shifts, fortified products, composite recipes, nondetects and mixed units. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Have an independent reviewer reconstruct the original data meaning. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Testing only ordinary raw foods misses the most consequential failures. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 20: Inspect controlled vocabularies

Purpose. Version tagnames, LanguaL facets, FoodEx2 terms and any local extensions. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Keep a resolver or codebook for every published mapping. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A code label may change without a code becoming a nutrient result. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 21: Reconcile duplicate names

Purpose. Detect translated names shared by distinct source foods or components. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Do not merge on strings; use qualified identifiers and versioned keys. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Name collision is a routine property of multilingual data. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 22: Roundtrip machine fields

Purpose. Export the target data and confirm source identifiers, units, values and statuses survive unchanged. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Use deterministic checksums or cell-wise comparison for locked fields. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A beautiful interface may conceal a broken CSV or JSON payload. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 23: Check accessibility

Purpose. Use expanded unit names where helpful, descriptive headers, readable tables and alternatives for symbols. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Retain exact data even when adding explanatory prose for accessibility. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Accessibility requires context, not lower measurement fidelity. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 24: Review expert exceptions

Purpose. Send uncertain scientific equivalences to a qualified food composition compiler or analyst. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Keep the exception ledger linked to the affected rows. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. Unresolved does not mean erroneous, but it must not be silently asserted as verified. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 25: Approve publication metadata

Purpose. Check title, scope, data source, methods, limitations and licence or reuse conditions. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Publish source citations and an explanation of derived fields. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A translated table without a citation trail is hard to reproduce and easy to misquote. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

Gate 26: Monitor changes after release

Purpose. Track amendments, source updates and corrected descriptors without rewriting historical evidence. This gate separates language choices from evidence decisions and creates a stopping point when required inputs are unavailable. It should be completed before a later gate attempts to compensate by inference.

Concrete deliverable. Issue versioned corrections or new derivatives with a changelog. The record should identify its owner, source material, decisions, exceptional cases and an independent way to test completion. Sign-off means the particular gate has been checked; it does not imply that all other dimensions of the table are scientifically validated.

Failure test. A past analysis may need its past data, even when a newer reference is better. Ask a second reviewer to reproduce the decision using the preserved source and documentation. If that reviewer needs unstated assumptions, return the item for clarification. Maintain both the machine-readable fields and a human-readable decision note so that a later team can distinguish a translated label from a newly computed datum.

7. Failure modes: why fluent nutrient translations can still be wrong

Failure 1: Food-name false friends

What goes wrong. Closely related words may refer to different species, dishes or processing states. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Require food source, edible form and state; do not use a lexical synonym as the final match. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 2: Name collision after translation

What goes wrong. Distinct records can acquire identical display names. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep stable source identifiers and a disambiguating qualifier in the interface. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 3: Raw/cooked reversal

What goes wrong. A prepared record is labelled as raw or vice versa. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Validate state against source fields and moisture or recipe documentation. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 4: Dry-matter basis lost

What goes wrong. A dry-basis concentration is displayed as an as-eaten value. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Preserve basis and convert only using a documented moisture figure. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 5: Per serving replaces per 100 g

What goes wrong. A portion result is copied into a reference table column. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Carry serving weight and calculation separately from source values. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 6: Milligram and microgram swap

What goes wrong. The magnitude changes by a factor of one thousand. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Check units independently of linguistic labels and add dimensional tests. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 7: Decimal comma parsing

What goes wrong. A decimal comma is treated as a thousands separator. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Store numeric values in a locale-neutral representation before formatting. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 8: Negative or less-than signs dropped

What goes wrong. A limit, correction or threshold loses its comparator. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Preserve signs and inequality symbols as structured data. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 9: Missing converted to zero

What goes wrong. Absent evidence is rendered as chemical absence. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Use explicit missingness codes and review zero-filling rules. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 10: Below-LOQ converted to nondetect

What goes wrong. A quantitative limit is presented as a categorical absence. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Retain comparator, threshold, method and reporting status. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 11: Trace converted to a quantity

What goes wrong. An unquantified trace is mapped to an arbitrary small number. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep trace as its documented data state. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 12: Carbohydrate method collapse

What goes wrong. By-difference and available carbohydrate fields are treated as the same. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Map only after checking method and fibre convention. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 13: Added sugar inferred from total sugar

What goes wrong. An analytical sugar measurement is taken as a formulation claim. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep composition and ingredient-origin claims separate. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 14: Protein factor silently changed

What goes wrong. Nitrogen-to-protein conversion is recomputed without documentation. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Preserve source factor and make recomputation a distinct output. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 15: Energy formula silently standardised

What goes wrong. A reference energy amount is replaced using another convention. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Retain source result plus an explicitly labelled derived calculation. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 16: Fibre methods flattened

What goes wrong. Method-specific fibre results are mapped to a single unqualified label. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep tagnames and method definitions as mapping conditions. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 17: Retinol treated as total vitamin A

What goes wrong. One chemical form becomes a total activity value. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Use form-specific terms and record equivalence factors. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 18: Folic acid mixed with folate

What goes wrong. Fortification and food folate are reduced to a single unqualified number. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Retain constituent forms and any dietary-equivalent calculation. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 19: Fatty-acid sum mislabelled

What goes wrong. A selected subset of fatty acids is called total fat. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Distinguish analyte sum from total lipid content. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 20: Fortification omitted

What goes wrong. Fortified and unfortified versions become indistinguishable. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Preserve enrichment status and formulation date. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 21: Product version ignored

What goes wrong. An older branded label is treated as a current recipe. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Record product and release identifiers with review dates. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 22: Sample mean treated as individual sample

What goes wrong. A statistical summary is presented as one measurement. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Retain summary type and sample count where available. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 23: Range flattened to exact value

What goes wrong. Variation across samples is concealed. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Report range, variance or uncertainty information as provided. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 24: Method code translated into prose

What goes wrong. A fixed machine code is replaced with a target-language word. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Translate its display description, never its stable identifier. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 25: Classification asserted as laboratory proof

What goes wrong. A FoodEx2 or LanguaL match is mistaken for measured composition. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Separate classification confidence from nutrient evidence quality. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 26: Edible fraction guessed

What goes wrong. A universal yield percentage is applied to different foods. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Require a food- and state-appropriate coefficient with a source. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 27: Cooking yield treated as retention

What goes wrong. Mass change is confused with nutrient survival. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep recipe mass balance and component retention in distinct calculations. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 28: Ingredient substitution masked

What goes wrong. A generic food match silently replaces a recipe ingredient. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Flag each substitution and re-estimate only with explicit approval. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 29: Salt conflated with sodium

What goes wrong. Masses are interchanged without a convention. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep sodium data original and label any salt equivalent as derived. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 30: Volume assumed equal to mass

What goes wrong. Liquid density is ignored. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Use a sourced density or retain the volume basis. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 31: Rounding creates phantom precision

What goes wrong. The translation implies more certain numbers than the source. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Match source significant digits and maintain raw values. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 32: Reference citations removed

What goes wrong. The translated table no longer links to its measurements. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep references, release and retrieval metadata attached to rows. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 33: Aggregate hides incompatible methods

What goes wrong. Two databases’ values are averaged despite distinct analytes. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Require equivalence rules before pooling; otherwise report side by side. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 34: Transliterated variety confused with species

What goes wrong. A food cultivar name is mistaken for a different crop. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Preserve scientific identification where supplied and show transliteration as metadata. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 35: Recipe translation changes instructions

What goes wrong. Preparation verbs alter nutrient retention assumptions. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Validate steps and recipe outputs independently. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 36: Allergen claim inferred from nutrient data

What goes wrong. A composition table is used as proof of allergen absence. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Refer to appropriate specialised evidence; keep this article focused on composition. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 37: Stale identifier mapping

What goes wrong. A vocabulary term changes between releases without an updated crosswalk. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Version mapping rules and rerun comparisons on change. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 38: Nutrient blank presented as compliant

What goes wrong. Absence of data becomes a safety or legal finding. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Do not use missing entries as policy conclusions. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 39: Sampling context generalised

What goes wrong. One cultivar, season or production system is projected onto all foods. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Keep representativeness limits visible and cite the sample context. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

Failure 40: Derived field lacks a calculation ledger

What goes wrong. A number is correct-looking but cannot be reproduced. The translated table may pass a casual readability check precisely because the labels sound familiar. The defect emerges when someone joins datasets, calculates dietary exposure, compares foods or quotes a number without looking up the source record.

Repair. Store source values, equation, factor, rounding and basis in the derived record. Before approving the correction, examine the original field, record ID, units, reference basis and data provenance. Document whether this was a linguistic misreading, an unsupported scientific mapping, a conversion error or an outdated source, because different causes need different preventive controls. A repair is complete only when the corrected record round-trips without losing its original evidence.

8. Worked diagnostic laboratory: fictional examples that test the whole translation chain

All numerical values, record fragments and foods in this laboratory are constructed teaching cases unless clearly marked as a description of a cited standard. They are not authoritative food composition values. Each case asks what a translator should preserve, which apparent shortcut is unsafe and how a second person can verify a correction.

Diagnostic 001: Cooked rice has a surprisingly low protein figure

Source fragment. 100 g cooked rice, 2.7 g protein; comparator 100 g dry rice, 7.5 g. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Claiming the cooking destroyed most protein per equal serving. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Compare like masses and disclose added water; do not calculate nutrient loss from concentration alone. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic questions. Would a user receive the same scientific answer if the source-language and target-language records were compared cell by cell? Is the corrected output reproducible from declared data rather than from the translator’s intuition? Does a search for the target food retrieve more than one plausible candidate? If so, write down the disambiguating attribute and make it visible. This case passes only when linguistic fluency and numerical fidelity are both preserved.

Diagnostic 002: A canned bean sodium discrepancy

Source fragment. One record reports drained beans, another beans with packing liquid. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Asserting an international labelling difference is the sole cause. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Check whether food fraction, rinsing and brine inclusion differ. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 003: A dried mushroom vitamin D spike

Source fragment. A dried sample reports vitamin D per 100 g with an exposure note. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Translating dried mushroom as fresh ordinary mushroom. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain drying and light-exposure conditions before any comparison. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 004: A cereal energy value changes after a software migration

Source fragment. Legacy energy field 1008 differs from a Foundation value associated with a 2047/2048 energy calculation. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Forcing the old energy column into a new code on label similarity. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Consult the FDC documentation; preserve source type, calculation and field identity. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 005: Fibre in a flour table does not match another database

Source fragment. Two sources use method-specific fibre definitions. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Reporting one value as erroneous because the other is larger. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Check official component definitions and laboratory methods. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 006: A small fish mineral figure looks impossible

Source fragment. Original concentration is in micrograms per 100 g, target table shows milligrams. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Revising the amount without auditing the unit. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Restore the source unit and use dimensional conversion with a logged equation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 007: A salt claim comes from sodium

Source fragment. Original source provides sodium in mg/100 g. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Publishing that number under an unchanged salt heading. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve sodium and label any sodium-to-salt equivalent as a separate calculation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 008: The component is below the quantification limit

Source fragment. Original laboratory result reports <0.05 mg/100 g. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Recording a factual zero to simplify charts. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain the comparator and report method detection limits where available. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 009: An empty folate cell becomes zero

Source fragment. A source row has no folate measurement. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Claiming the food contains none. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Encode missing and exclude it from calculations that require a measured amount. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 010: A recipe appears to create calcium

Source fragment. Final dish has lower mass after baking and a higher calcium concentration per 100 g. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Inferring calcium was chemically generated by heating. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use mass balance, edible yield and ingredient contributions before asserting changes. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 011: A boiled vegetable appears to lose protein

Source fragment. Values are reported per 100 g before and after water absorption. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating concentration change as equal to nutrient destruction. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Inspect preparation, final yield and absolute nutrient mass. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 012: An egg row has a cholesterol disagreement

Source fragment. One table covers whole egg, another egg white. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Calling the values conflicting references to the same food. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain biological fraction and reject the false match. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 013: A branded cereal changed its vitamins

Source fragment. The manufacturer reformulated and the source dates differ. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Overwriting the historical product with the newest label. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep both formulations as versioned records and cite the dates. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 014: A food name has no direct equivalent

Source fragment. A traditional mixed dish uses an unfamiliar source-language name. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Substituting the closest familiar prepared dish. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep transliteration and describe actual ingredients without claiming equivalence. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 015: Cooking yields are inverted

Source fragment. Recipe starts at 600 g and ends at 450 g; the processor uses 600/450 as if it were final mass yield. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Multiplying per-100-g values in the wrong direction. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Define yield factor as final divided by starting mass and show the calculation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 016: A source table reports potassium as trace

Source fragment. The status label is text rather than a number. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Using zero in the target table to permit numeric sorting. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain the status with a nonnumeric or explicitly censored value. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 017: Vitamin A units have the same displayed label

Source fragment. Two source values report distinct activity-equivalence conventions. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Aggregating them into one ordinary microgram column. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve basis and present incomparable values separately until converted explicitly. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 018: Iron is listed for fortified flour

Source fragment. The source identifies enrichment. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating the value as representative of every wheat flour. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Carry fortification and formulation qualifiers into the food label. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 019: A dry seaweed figure appears enormous

Source fragment. 100 g of dried seaweed is compared with 100 g of hydrated product. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Presenting the dried concentration as a prepared serving. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain moisture basis and use actual serving information only if documented. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 020: A tea extract is mistaken for whole leaf

Source fragment. The record describes brewed infusion. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating all leaf components as fully extracted into water. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Distinguish source ingredient and beverage product; never invent an extraction percentage. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 021: The nutrient table uses a comma decimal

Source fragment. 1,25 in the source locale means one and a quarter. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Parsing it as 125 or one thousand twenty-five. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use locale-aware parsing and store a machine-safe numeric value. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 022: An experimental food has a notable value

Source fragment. A study reports a particular experimental cultivar under specific conditions. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Using it as a universally representative food value. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain experiment context and data type from the source. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 023: A consumer label and laboratory table disagree

Source fragment. A branded-food record derives from manufacturer data, a Foundation record from analytical sampling. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Calling one value a translation error on numeric difference alone. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Compare data type, product identity, scope and rounding before judgment. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 024: A mixture includes oil that the name omits

Source fragment. A ready-to-eat vegetable dish contains added oil. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Matching its nutrient value to plain boiled vegetables. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve recipe ingredients and mark the food as composite. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 025: An apple portion includes inedible core

Source fragment. Purchase weight is 160 g but edible mass has not been reported. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Multiplying a per-100-g edible value by the full purchase weight. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Request a sourced refuse or edible portion estimate. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 026: Protein from nitrogen uses a special factor

Source fragment. Total nitrogen 2.0 g/100 g is combined with an explicitly reported factor. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Replacing the factor silently with a standard default. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep original calculated protein and disclose any alternative computation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 027: Carbohydrate by difference is read as sugars

Source fragment. A total-by-difference column is translated as sugar. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Misrepresenting starch and other fractions as simple sugar. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Restore component identity and use the source’s sugar field if available. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 028: Yogurt brand name implies natural but is sweetened

Source fragment. Product formulation includes added sweetener. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Removing the addition because the target descriptor sounds cleaner. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use factual ingredient and product data, not marketing connotations. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 029: Water and oil packing share the same fish label

Source fragment. Both candidates have a generic tuna translation. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Joining records by translated name and averaging fat. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Add packing medium and drained status to the join key. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 030: A frozen meal has two values per serving

Source fragment. One version refers to frozen as sold, another to prepared following directions. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating their denominators as identical. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain preparation instructions, final weight and data type. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 031: Folic acid and food folate appear combined

Source fragment. A fortified food reports multiple folate-related fields. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Flattening all forms to one unlabeled vitamin name. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve folic acid, food folate and DFE distinctions as sourced. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 032: The classification code is not a nutrient ID

Source fragment. A FoodEx2 base term is attached to a food sample. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Placing the code into a food component column. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep food classification and component identity in separate metadata fields. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 033: A LanguaL facet describes preservation

Source fragment. A coded facet indicates drying or freezing. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating the facet as direct evidence of final moisture percentage. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve the classification and demand an actual moisture value for calculations. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 034: A nutrient mean hides variation

Source fragment. A Foundation food record aggregates samples collected over time. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Quoting the mean as a guaranteed value for every sample. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Report sample context and any available spread as part of the evidence. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 035: A serving is expressed as a cup

Source fragment. The portion database gives a specific gram weight for that food. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Assuming all cups are 240 g. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use the product-specific weight instead of a generic mass-volume identity. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 036: A fortified plant beverage is matched to plain soybeans

Source fragment. The beverage contains water and additives. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Interpreting beverage calcium as intrinsic to the seed. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve processing and fortification; disallow the join. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 037: A nut butter is compared with raw nuts

Source fragment. The spread may contain sugar, salt and added oil. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Using a shared nut name as proof of component equivalence. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Separate ingredient and formulation data. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 038: A cooked meat row excludes fat

Source fragment. The label says trimmed after cooking. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Translating it as a whole edible meat portion. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Carry trimming stage and fat status into the target. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 039: Shrimp protein is calculated on shell-on mass

Source fragment. The source composition refers to edible flesh. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Using total purchase weight as the nutrient denominator. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Apply a sourced edible yield or report that conversion unavailable. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 040: A spicy dried seasoning looks rich in nutrients

Source fragment. The comparison is conducted on a 100 g basis. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Presenting that value as what a normal dish supplies. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain analytical concentration and separately report documented usage amounts. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 041: An old database code survives in a new export

Source fragment. A legacy component code maps only partly to the new convention. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Blindly renaming the code because the displayed label resembles it. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep original code, version and a qualified crosswalk decision. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 042: A label value has been rounded

Source fragment. Manufacturer states a whole-number value, analytic source has decimals. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Claiming an exact quantitative discrepancy from displayed digits alone. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Check reporting and rounding rules before asserting divergence. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 043: An imputed micronutrient value is displayed as observed

Source fragment. A compiler borrowed a value from a comparable food. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Publishing it without the imputation flag. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Carry estimation provenance and separate it from measured samples. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 044: An official food classification term is revised

Source fragment. A FoodEx2 update changes term availability or hierarchy. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Reinterpreting historic records using the new hierarchy without noting it. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain classification release and migrate with a documented mapping. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 045: Vitamin E field is overly broad

Source fragment. The source specifically measures alpha-tocopherol. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Translating as total tocopherols. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve analyte form and method. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 046: Omega-3 totals aggregate unlike fatty acids

Source fragment. A table sums ALA, EPA and DHA while another lists only EPA+DHA. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Calling them identical omega-3 values. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Carry the exact constituent list and refuse an unqualified merge. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 047: The source has both sodium and salt equivalents

Source fragment. One figure is measured or compiled sodium, the other derived. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Doubling the nutrient by counting both as independent constituents. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Label each role and prevent duplicate contribution. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 048: A sourdough bread row is not plain bread

Source fragment. Fermentation and ingredients are recorded in the source. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Substituting the common translated bread name. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep processing and ingredients without inventing a fermentation effect. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 049: A high iodine value belongs to dried product

Source fragment. The source sampled dry seaweed. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Calling it a ready-to-eat hydrated serving. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Maintain dry basis and give no unsourced dietary recommendation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 050: A mineral-water value is per litre

Source fragment. The export screen shows mg/L. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Displaying it as mg/100 g without a valid conversion. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep original concentration basis and only convert with appropriate density assumptions. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 051: A calculated energy value has a different fibre term

Source fragment. One convention attributes energy to fibre and another handles it differently. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Claiming arithmetic mismatch proves corruption. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Inspect the source factors and carbohydrate definition. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 052: A trace fatty acid gets rounded away

Source fragment. The source says trace or <LOQ for a fatty acid. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Using a dash that readers interpret as zero. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Explain status in the table legend and preserve original marker. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 053: The source recipe uses cooked ingredient weights

Source fragment. The translation mistakenly treats weights as raw. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Recomputing a soup from wrong input masses. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve ingredient state for each line item. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 054: A prepared formula row uses powder data

Source fragment. The source gives dry formula concentration per 100 g. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Publishing it as 100 mL of prepared feed. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Stop publication of the converted field until preparation and relevant review are established. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 055: A food descriptor is a brand slogan

Source fragment. The source uses a promotional phrase not a standard classification. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating the slogan as an analytical statement. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use neutral description and retain trademark as an identity field if appropriate. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 056: The exact food variety is unknown

Source fragment. A generic fruit category lacks cultivar or sampling information. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Inventing a variety because it is common in the target language. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep generic source scope and record uncertainty. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 057: A missing reference makes values hard to audit

Source fragment. A row includes nutrient amounts but lacks citation or data type. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Presenting it as verified laboratory work. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Require lineage or mark the evidence as unsupported. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 058: A food table merges incompatible years

Source fragment. Nutrient values from releases years apart are silently consolidated. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Claiming the composite table is a translation of one source. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Identify every source contribution and the chosen merge rule. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 059: Dry matter and wet weight are both reported

Source fragment. A study provides two legitimate expressions of the same analyte. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Averaging them as two independent observations. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Recognise the basis conversion and do not double-count. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 060: A protein isolate has an unfamiliar translated name

Source fragment. The target language lacks a standard term for the extraction stage. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Replacing it with ordinary flour. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use a descriptive qualifier and preserve original process name. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 061: A traditional recipe has optional ingredients

Source fragment. Source documentation marks a spice optional. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Translating optionality as required and changing the implied nutrient calculation. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep the modality and make variant recipes separate if calculated. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 062: A freeze-dried berry record lacks rehydration

Source fragment. Nutrients are per 100 g powder. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Presenting values for reconstituted fruit without water amount. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Retain powder basis and flag requested dilution as unresolved. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 063: A sampling region is mistranslated as an origin guarantee

Source fragment. The record states where samples were obtained, not protected product origin. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Turning collection geography into a certification claim. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep sampling and product identity separate. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 064: An oil blend percentage is unknown

Source fragment. A record identifies a mixed oil product. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Assigning the same profile as one predominant oil. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve blend status and avoid unlicensed component averaging. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 065: An unknown cereal ingredient is assumed gluten-free

Source fragment. The table’s nutrient record lacks allergen evidence. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Claiming absence of gluten from a generic food name. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Separate nutritional composition from allergen validation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 066: The source used an older iodine analytical method

Source fragment. Methods differ across releases. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Declaring an apparent increase to be purely dietary. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Carry method identifiers and assess comparability before an interpretation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 067: A flavoured drink uses liquid volume for servings

Source fragment. The nutritional concentration is reported per 100 mL. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Multiplying by a serving size given in grams as though density equals one. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Keep volume and mass distinct or use sourced density. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 068: A fortified product loses its source version in a translation

Source fragment. A formulation was updated mid-year. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Publishing one unqualified cross-language table for both versions. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Pin the source release and show versioned comparisons. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 069: A recipe retention factor is applied twice

Source fragment. Ingredient nutrients were already based on cooked entries. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Applying a second cooking loss factor to the compiled dish. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Trace each ingredient to prevent duplicated transformation. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 070: An analyte is expressed as a chemical equivalent

Source fragment. A vitamin value reports an activity-equivalent rather than mass of one molecule. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating the equivalent unit as direct chemical mass. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Preserve convention and derive only with a documented formula. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 071: A table uses an uncertain food match

Source fragment. The target ingredient is only approximately comparable to a database item. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Presenting the mapping as exact with no note. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Record match confidence, attributes and alternative candidates. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 072: A dietary survey code is mistaken for an FDC identifier

Source fragment. Survey categories and FoodData Central records have different roles. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Joining solely on a numeric-looking code. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use source-specific identifier namespaces and explicit mapping tables. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 073: A calculated recipe leaves out the cooking liquid

Source fragment. The cooking liquid is part of the consumed dish. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Applying a drained-food assumption. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Match final consumed recipe and water balance. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 074: Several duplicate nutrient names occur in one export

Source fragment. Two columns have identical translated labels but different method codes. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Dropping one as a duplicate. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Use component IDs and method qualifiers to avoid destructive deduplication. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

Diagnostic 075: A product is listed as ready-to-eat but requires dilution

Source fragment. The source differentiates concentrate from consumed form. The exercise assumes that the number or phrase belongs to a versioned food-composition record. First reconstruct the full proposition: food identity and state, component definition, amount, unit, denominator, analytical or calculated status, and the cited source. A name alone cannot settle which record is intended.

Tempting but invalid translation. Treating marketing shorthand as preparation evidence. This can mislead because a natural-language reader may attribute a stronger claim to the table than the data justify. Mark the exact boundary that has been crossed: food matching, measurement basis, biochemical form, data provenance, inference or presentation.

Defensible correction. Request preparation details before calculating intake amounts. Keep source identifiers and the original expression in the audit ledger. Publish the translated label only with the qualifier needed to reconstruct the source, and create a separate derived field if conversion or estimation is necessary. An independent reviewer should be able to state what remained unchanged, what was transformed, and which assumption still requires evidence.

9. Transfer workshops: apply the method without inventing data

Workshop 01: Build a two-language food identity card

Task. Take an unfamiliar food name and produce source label, target gloss, scientific or product identity, edible portion, processing, preparation and a list of unresolved attributes. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. A defensible card makes the food searchable without declaring a false equivalent to a familiar ingredient. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Transfer test. Repeat the exercise with a second food category or a different nutrient definition. The method should survive the change of domain without treating names as primary keys or flattening missing values, measurement procedures or processing states. Keep a short record of what had to be changed and what remained invariant.

Workshop 02: Reconstruct the nutrient assertion

Task. From a table cell, record food, component, value, unit, basis, method/status and source. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. A passing reconstruction leaves no hidden denominator or missing provenance. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 03: Separate measured from calculated protein

Task. Given nitrogen and an explicit conversion factor, retain the original value and show any arithmetic as a derived field. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. A different factor must be disclosed instead of silently substituted. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 04: Audit energy conventions

Task. Compare a source energy value with a transparent calculation based on its stated macronutrient definitions. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Document why differences may arise before calling either value erroneous. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 05: Preserve a less-than result

Task. Translate a <LOQ-style numeric field and explain what the threshold does and does not establish. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Do not emit a numeric zero without a stated imputation policy. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 06: Create a LanguaL mapping sheet

Task. Map food attributes to approved facets while keeping machine codes intact. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Record the vocabulary version and distinguish description from analytical proof. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 07: Explain a FoodEx2 choice

Task. Write a food category assignment with base term, applicable facets and reviewer confidence. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Make a classification choice auditable rather than relying on a translated name. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 08: Version a branded food record

Task. Compare two formulation dates and their source nutrient displays. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Retain historical versions and note that a changed label is not necessarily a translation problem. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 09: Reverse an accidental mg-to-µg error

Task. Use dimensional analysis to repair a thousandfold display defect. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Confirm the original unit and value before issuing any converted table. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 10: Review a dry-versus-wet mismatch

Task. Identify which water basis each data source uses. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Avoid drawing conclusions about absolute nutrient retention from concentration alone. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 11: Calculate an edible portion example

Task. Given a hypothetical purchased mass and sourced edible fraction, calculate the edible mass explicitly. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Label the fraction’s provenance and applicability to the food’s preparation. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 12: Rebuild a recipe mass balance

Task. Record starting ingredient masses, final cooked mass and the assumed cooking yield. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Keep mass change distinct from nutrient retention. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 13: Apply component-specific retention

Task. Given a documented nutrient retention factor and ingredient nutrient mass, show a separate derived calculation. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Do not apply one universal factor to all nutrients or double count cooking transformations. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 14: Diagnose carbohydrate method conflict

Task. Compare carbohydrate by difference with analytically assessed available carbohydrate. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Retain both definitions and avoid pretending they are interchangeable. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 15: Create a folate forms panel

Task. Label folic acid, food folate and any dietary equivalent with explicit units. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Identify source method and formula before combining forms. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 16: Flag apparent sample variation

Task. Read mean, sample count, range and date when supplied. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. A generalised summary must not be presented as the guaranteed composition of an individual item. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 17: Validate source citations

Task. Choose a food row and find its original source and database release. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. If lineage is absent, report the limitation rather than fabricating a laboratory reference. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 18: Design an imputation disclosure

Task. For one missing nutrient, propose an explicitly labelled inference method and reason to reject it. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. The output should preserve the missing original and the estimated derivative separately. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 19: Test an ambiguous culinary name

Task. Translate a recipe name with two plausible target-language dishes. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Retain ingredients, qualifiers and non-equivalence notes until a justified match exists. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 20: Audit a unit display layer

Task. Render values in languages using different decimal separators and thousands conventions. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Roundtrip every stored number to ensure no magnitude change. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 21: Create an accessibility-first table

Task. Add readable headers, explanatory notes and text alternatives for mathematical symbols. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. The improved display must still carry exact values and status indicators. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 22: Build a change log

Task. Record why a component code, food term or data mapping was amended. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Do not overwrite the history of previous analyses that depended on the old rule. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 23: Reconcile two data sources

Task. Compare a Foundation Food, a branded label and a dietary survey food in terms of provenance. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Explain why identical nutrient column names do not make data sources identical. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

Workshop 24: Write a reviewer sign-off

Task. Produce separate linguistic, numeric and provenance review decisions for one complex row. Begin with the original record and its provenance, then write a source-locked field inventory. Produce a target-language display that a reader can understand without overwriting the source values or codes. Record any interpretation or conversion on a separate derived line, with the date and a named review responsibility.

Evidence of success. Show unresolved issues rather than compressing all review into one pass/fail score. For peer review, exchange outputs with someone who did not see your reasoning. Ask the reviewer to recover the original object, its unit and measurement basis, methodological qualifier, and the circumstances under which the translated statement would be false. If the reviewer cannot reconstruct those boundaries, improve the data structure rather than adding promotional language or general assurances.

10. Working glossary: what each term commits you to

1. Food composition database

A curated collection of food identities and component values with supporting documentation. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

2. INFOODS

The FAO-coordinated International Network of Food Data Systems focused on composition quality and exchange. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

3. Component identifier

A stable code for a defined nutrient, chemical component or related measured property. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

4. Tagname

An INFOODS-style named identifier for a food component or calculation factor. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

5. Food matching

Choosing a relevant food record to represent a described food under explicit criteria. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

6. Food nomenclature

The naming and distinction of food types, parts, preparations and products. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

7. Edible portion

The portion of a food intended for consumption, distinct from refuse or non-edible matter. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

8. Refuse

Material omitted from the edible composition basis, such as specified shells, pits or bones. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

9. Dry-matter basis

A way of expressing composition with water excluded from the mass denominator. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

10. As-consumed basis

The preparation state intended to represent the food when eaten. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

11. As-sold basis

The condition in which a packaged product is sold, not necessarily ready for consumption. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

12. Processing state

The documented treatment or manufacturing condition of a food. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

13. Preparation state

How a food was made ready for measurement or consumption. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

14. Food source

The underlying plant, animal, fungal or other material from which a food derives. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

15. Cultivar

A cultivated variety or selection, relevant where documented. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

16. Recipe

A defined mixture and preparation of ingredients used for an aggregate composition calculation. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

17. Recipe yield

The relationship between initial ingredient mass and final prepared dish mass. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

18. Nutrient retention

A measure or factor associated with the amount of a nutrient remaining after a process. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

19. Moisture change

Water gain or loss that alters concentration per unit food mass. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

20. Measured value

A value produced from an identified measurement method and sample. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

21. Calculated value

A value obtained from an equation or specified inputs rather than a direct assay of that result. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

22. Imputed value

A filled data gap produced by a declared estimation method. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

23. Estimated value

A value inferred under stated assumptions rather than directly measured for that precise record. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

24. Borrowed value

A value carried from another source food record with a documented matching rationale. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

25. Analytical method

The laboratory procedure used to determine a component. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

26. Method dependency

A difference in what a value means that follows from its measurement or calculation method. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

27. Reference basis

The denominator and state relative to which a value is expressed. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

28. Concentration

Amount of a component per stated unit of food mass or volume. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

29. Per 100 g

A reference amount that requires a defined edible object and preparation state. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

30. Serving

A declared or inferred amount of food; it should have a documented mass or volume. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

31. Portion weight

The mass associated with a specified portion description. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

32. Density

Mass per volume, needed for certain valid mass-volume conversions. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

33. Significant digits

Displayed precision that should not claim more certainty than the source supports. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

34. Limit of detection

A method-defined threshold concerning detectability under stated conditions. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

35. Limit of quantification

A method-defined boundary for quantitative reporting under stated conditions. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

36. Below limit

A censored result that should retain its comparator and reference limit. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

37. Trace

A source-specific data status that does not automatically supply a numeric quantity. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

38. Missing value

An absent or unavailable result rather than a measured zero. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

39. Not applicable

A value state meaning a field is not relevant under the declared model. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

40. Sampling

The selection or collection of material used to generate a composition record. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

41. Representative data

Values intended to describe a relevant food population, subject to sampling limitations. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

42. Sample mean

An average over a documented set of observations. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

43. Variance

A measure of spread associated with observations. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

44. Provenance

The evidence trail from source and methods to current data. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

45. Release

A documented published version of a dataset or code system. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

46. Record identifier

A key assigned to an item in its source system. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

47. Canonical key

An internally stable machine representation not replaced by language display. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

48. Crosswalk

A mapping table between two systems’ records or component codes. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

49. One-to-one mapping

An asserted correspondence that must be justified by matching definitions. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

50. Approximate match

A candidate substitution with acknowledged incomplete equivalence. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

51. Controlled vocabulary

An organised set of permitted terms or codes with defined meanings. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

52. LanguaL

A structured food description vocabulary using facets. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

53. FoodEx2

EFSA’s food classification and description framework with hierarchies and facets. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

54. Faceted classification

Combining multiple descriptors to represent aspects of one entity. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

55. FoodData Central

USDA’s collection of distinct food-composition data types. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

56. Foundation Foods

FDC data type with analytical values and detailed sample metadata. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

57. Branded Foods

FDC data type drawn from manufacturer-provided product label information. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

58. FNDDS

Food and Nutrient Database for Dietary Studies, used with dietary intake surveys. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

59. SR Legacy

A historical FDC data type with its final release in April 2018. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

60. Experimental Foods

FDC data type including study-related foods and associated research context. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

61. EuroFIR

A food-composition harmonisation network and resource associated with common data standards. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

62. Energy factor

A specified conversion used to compute energy contributions from components. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

63. Carbohydrate by difference

A calculated quantity using subtractions from a reference mass. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

64. Available carbohydrate

A component definition that requires a stated analytical or calculation convention. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

65. Total sugar

A compositional measure not synonymous with sugars added during manufacture. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

66. Dietary fibre

A method-dependent food component often distinguished using controlled identifiers. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

67. Folate equivalent

A derived vitamin-related measure whose conversion basis must be stated. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

68. Retinol activity equivalent

A defined vitamin A activity expression not identical to retinol mass. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

69. Vitamin form

A particular chemical or activity-defined member of a nutrient family. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

70. Nutrient form

The specific measured or calculated component behind a familiar nutrient label. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

71. Data type

A category of records differing by source, scope and methods. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

72. Linguistic review

A check that target-language wording preserves the source’s asserted meaning. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

73. Data QA

Checks on amount, units, identifiers, methods, status and lineage. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

74. Roundtrip test

A check that export/import or translation preserves source-locked fields. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

75. Exception ledger

A record of unresolved, rejected or conditionally approved mappings. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

76. Derived field

A new value calculated or inferred from preserved source values under declared assumptions. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

77. Reproducibility

The ability to reconstruct an output from its documented source and method. In translation work, this term should be linked to the exact field or decision it describes. Do not replace its technical function with a looser ordinary-language synonym when the reader could then confuse a food, a component, a method or a result.

11. Frequently asked questions

1. Can I translate every local food name to the nearest familiar English dish?

No. A familiar dish name can imply ingredients, processes or edible portions not present in the source. Keep the original or transliterated identity and use a descriptive gloss when no documented equivalent exists.

Practical next move: locate the source field and its documentation, state the exact meaning in one complete sentence, and test whether the proposed target wording leads to the same inference. If an answer requires a chemical, statistical, recipe or policy assumption that the source does not provide, record that assumption in a separate reviewer note instead of presenting it as part of the translation.

2. What is the first check before moving a nutrient value to another table?

Confirm the food and component identities, unit, denominator, source data type and provenance. The apparent numeric match is less important than proving that both columns make the same kind of claim.

3. What is the purpose of an INFOODS tagname?

It identifies a defined component for data exchange and helps distinguish component concepts. Do not invent a tagname from a translated English label; consult the official list and its version.

4. Is LanguaL the same as FoodEx2?

No. They are separate structured vocabularies with different roles. Both can support food description, but their codes and hierarchy semantics should not be swapped without an explicit mapping.

5. Why distinguish Foundation Foods from Branded Foods?

The former emphasises analytical values and sample metadata; the latter includes manufacturer-provided label information. A translator should not imply that the records were produced by the same evidence process.

6. Can a missing value be treated as zero?

Not by translation alone. Missing, not analysed, below limit, trace and measured zero are distinct data states. A study may make an imputation choice, but it must be explicit and preserve the original missingness.

7. How do I translate a value below the limit of quantification?

Keep the inequality symbol, threshold, unit and methodological explanation. Do not replace it with a categorical claim of absolute absence.

8. What happens when a component name has no clean target-language equivalent?

Use the closest accurate descriptive label plus the original identifier and explanatory qualifier. A longer but precise phrase is preferable to a short false synonym.

9. May I change a source value to comply with a preferred energy formula?

You may calculate a separately labelled derivative if the purpose permits it and assumptions are documented. Do not present that derivative as the untouched translated original.

10. Why do raw and cooked values differ so much?

Preparation often changes water content and may change nutrient amounts. Values per 100 g can differ because the denominator changed; they do not automatically measure absolute nutrient retention.

11. What is a recipe yield factor?

It relates starting ingredient mass to final prepared food mass under a stated convention. It is not the same as the amount of each nutrient remaining after cooking.

12. Why does edible portion matter?

Composition values are frequently expressed for edible material, whereas purchase or household weights can include inedible fractions. Using the wrong mass changes computed nutrient quantities.

13. Can I use a universal conversion from cups to grams?

No. A cup describes volume; the corresponding mass depends on food, preparation and packing. Use a documented food-specific gram weight or retain the volume reference.

14. What is a method-dependent fibre problem?

Different analytical methods may capture different fractions. A shared fibre label does not prove quantitative equivalence; preserve method information and compare definitions.

15. Are total sugars and added sugars interchangeable?

No. Total sugars describes a compositional quantity under a defined method; added sugars relates to ingredients or formulation. Neither can be inferred reliably from the other’s label alone.

16. Is sodium identical to salt?

No. Sodium is an element measured or reported as a component; salt equivalents are derived under a convention. Use separate labels and documented calculations.

17. How should I handle vitamin A values?

Identify whether the source reports retinol, carotenoids, retinol equivalents, retinol activity equivalents or another defined expression. Carry the form and units before comparison.

18. What about folate and folic acid?

They are not automatically interchangeable. For equivalent calculations, use the source’s stated formula and keep chemical or activity-related terms distinct.

19. What information should accompany a translated nutrient table?

At minimum include source identity, food and component codes, preparation, units and bases, method or status where relevant, release, date, citations and clear flags for derived or unresolved values.

20. Can a machine translation system perform reliable mapping by itself?

It may propose candidate labels, but correctness depends on source definitions, numerical validation and domain-specific review. Fluent output should not be mistaken for verified scientific equivalence.

21. How do I prevent translated names from merging distinct foods?

Use stable source identifiers as keys and store translated names as display fields. Where two foods share a translated name, add qualified descriptors rather than deleting records.

22. Is every international food composition database directly comparable?

No. Foods, sampling populations, methods, data types, component definitions and source years differ. Harmonisation requires explicit criteria and may leave incompatible values separate.

23. Why keep historical releases?

They support the reproducibility of studies and published tables that used earlier data. Quietly changing values under the same translated name alters evidence retroactively.

24. Can a classification code replace a nutrient analysis?

No. Classification identifies what kind of food is described; it does not itself provide a measured concentration. Preserve codes separately from quantitative values.

25. How should a translator report uncertainty?

Carry any original statistical spread, qualifiers, limits, sample information and unresolved matching decisions. Do not replace them with confident prose or unsupported precision.

26. What is the safest default for an uncertain match?

Retain the original food identity and mark the target candidate provisional. Prevent automated merging until the attributes that matter are checked.

27. Does a good translation make the data medically suitable?

No. Accurate representation does not constitute individual medical advice or regulatory approval. Those uses require further professional assessment and context.

28. How do I review food matching at scale?

Build rule-based checks for identifiers, food state, unit consistency and data status, then independently sample difficult cases. Record ambiguous matches and audit them separately.

29. How do I cite a translated FoodData Central record?

Use the dataset’s source, record identifier where appropriate, data type, release or access date and method notes, following the source’s citation instructions. The translation’s own version should also be traceable.

30. What if two values differ because of a method change?

Document the method distinction. Do not force numerical agreement through language editing. If a harmonised derivative is justified, label its assumptions and source separately.

31. Can a glossary replace the source codebook?

No. A glossary explains terms for readers; a codebook defines machine fields and methodological constraints. A sound workflow uses both without allowing the glossary to alter identifiers.

32. What makes a correction transparent?

It identifies the original problem, the exact affected fields, the source used to resolve it, the corrective transformation and the version in which the fix appears.

33. How should I choose between a general label and technical precision?

Use a readable label when meaning is preserved; add a definition or qualifier when one ordinary word would hide distinctions. Keep machine-level precision regardless of display choice.

34. How should I handle data from a sample whose origin differs from the target user’s food?

Preserve source sampling context and explicitly flag representativeness limits. Do not silently claim it describes every variety, season, location or formulation.

35. What is the stopping rule for publication?

Do not release a numeric claim that has lost its unit, reference basis, source identity, component definition or unresolved critical qualifier. Publish a qualified partial description rather than an unsupported exact value.

36. How does this article connect to general translation learning?

It applies the general principle of moving meaning without moving evidence. Readers can use the series root for general methods, and consult separate vocabulary or English-system hubs for language skills rather than treating this specialism as their owner.

12. A four-week implementation plan

Week 1: learn the objects

Select ten food records spanning raw ingredients, prepared foods, commercial products and recipes. Build a field inventory for every record: identity, food state, component, value, unit, denominator, method, status, source, version and references. Compare how a source displays the information in human-facing pages and in downloadable data. Make no conversions until the records can be reconstructed. The outcome is a bilingual food identity sheet with source-locked keys and clearly marked uncertainties.

End-of-week check: explain one error that became visible only after checking a different evidence layer. Demonstrate the fix with before-and-after field representations, then show that the final human-language description and the source-locked numeric record both remain truthful. Carry unresolved questions into a versioned ledger instead of assuming that time spent editing has resolved them.

Week 2: learn component definitions

Choose twelve nutrient fields, including energy, protein, carbohydrate, fibre, a mineral and several vitamin forms. Find official tagnames or database definitions where available. Write a target-language label and a method note for each; record cases where a direct mapping is unsafe. Test missingness, less-than values, decimals and unit signs. The outcome is a component crosswalk that distinguishes approved matches, conditional mappings and incompatible concepts.

Week 3: learn transformations

Work through edible portion, portion weights, cooking yield, retention factors and selected recipe examples using fictional values or clearly documented public reference data. Keep source and derived columns separate. Run roundtrip tests so exported values, IDs, unit and status survive unchanged. Have another reviewer check all formulas, source states and assumptions. The outcome is an auditable transformation ledger, not merely a translated spreadsheet.

Week 4: learn publication and correction

Create a compact multilingual sample table and supporting methods page. Include source version, definition notes, limitations and an exception ledger. Ask independent reviewers to reconstruct at least ten difficult rows; revise any wording that causes a false inference. Record every correction and freeze a release. The outcome is a demonstrably reproducible data translation with a clear reader route to the original sources and an honest description of what remains uncertain.

13. Research sources, terminology authorities and scope boundaries

The terminology and method distinctions in this reference were checked against these authoritative public resources. For actual production decisions, consult the current version of the relevant documentation and the original dataset’s release-specific codebook. Some resources describe related but distinct data ecosystems, not one universal standard.

FAO Food Composition / INFOODS: the international food composition programme and resource network.
FAO/INFOODS Standards and Guidelines: food nomenclature, component identifiers, food matching and data checking guidance.
FAO Pacific Islands Food Composition Tables: Components and Factors: documented component names, energy factors and unit examples; use edition-specific methods rather than generalising them to every dataset.
EuroFIR Quality and Standards: compilation SOPs, LanguaL descriptions, source quality, recipe calculations and retention.
EuroFIR Food Identification and Description: controlled food description and indexing context.
USDA FoodData Central Data Type Documentation: Foundation, Branded, FNDDS, Experimental and SR Legacy distinctions.
USDA Foundation Foods Documentation: sample detail, component calculation, portions, energy and LOQ handling.
EFSA Data Standardisation, reviewed 23 September 2026: FoodEx2 hierarchies and facets and separate SSD2 data conventions.
EFSA FoodEx2 Maintenance 2025, published 9 April 2026: terminology updates and version-aware classification.
ISO 5060:2024: guidance for human evaluation of translation output, including error analysis and sampling.

Source-use caution: these links establish terminology, official data categories and documented concepts. They do not make the fictitious example numbers in this page reference dietary values. Official food composition records must be consulted for real calculations. A laboratory analytical method, a food description classification and an assessment framework each govern different decisions. This article does not certify any table for clinical, legal, labelling or population assessment use.

14. Next routes: build from a specialised data problem to durable translation skills

Return to the parent Master Art of Translation | The Complete System for Moving Meaning Between Languages when you need the shared model for intent, terminology, evidence, quality assurance and transfer across domains. This food-composition child adds the food identity, component method, reference-basis and provenance rules that the parent deliberately does not claim to exhaust.

For systematic learning of word meaning and precise distinctions without expanding this article into an unrelated course, visit the protected Vocabulary Learning Hub. For syntax, explanation, context and faithful English expression in a separate owner, use How English Works | The Complete Map and Learning Routes. These are outward learning routes, not permission to replace or rewrite their pages.

Final decision rule: publish a translated food-composition assertion only when its source food, component definition, value, unit, basis, status and evidence trail can be recovered. If any decisive field is unresolved, show the limitation. An uncertain but traceable record is more useful than a confident translation that changes the food or the number.