How to Translate Names and Preserve Entity Identity

How to translate names and preserve identity across languages

How do you translate proper names, personal names and place names into another language without changing who or what they identify? A dependable method separates the real-world entity from its written labels, approved names, transliterations, script variants and display conventions. Translation may change the visible form of a name while its identity must remain stable. This guide explains name translation versus transliteration, cross-script name matching, multilingual entity identity, aliases and practical quality checks for readers, translators and international publishing teams.

The hidden problem is that identical spelling is not proof of identical identity, and different spellings are not proof of different identity. One person may have an original-script name, a preferred Latin-script form, a public professional name and a historical record under an earlier name. Two unrelated people may also have the same display name. A translated article that merges them, invents a surname, mistranslates a brand, substitutes a nearby place or changes a specimen identifier may remain grammatically fluent while asserting something materially false.

People searching for how to translate names correctly, when to transliterate instead of translate, how to keep names consistent across languages or how to handle aliases in multilingual documents need a repeatable decision system. Begin with a source record and stable entity key, preserve the original name and approved variants, select the proper display convention for each target context, validate any transformation, and test whether a reader can trace the translated mention back to the intended entity. This article builds that system through mechanisms, failure analyses, diagnostics, worked cases and transfer exercises.

The job is not to replace a dictionary of names, an introduction to onomastics, a multilingual search algorithm or a grammar lesson on pronouns. It is to preserve entity identity during the movement of meaning between languages and media. A name may have an established target-language form, but only evidence or a controlled policy should justify choosing it. The identity must not change when a script, title, order, abbreviation, quotation or metadata field changes.

All individual and business examples introduced as practice cases are fictional unless explicitly identified as common standards examples. Case studies are training situations, not statements about actual people, registered entities or verified records. Legal names, civil records, personal identity documents, medicine, banking and official registries require the applicable authority’s rules and qualified review. Translators should not invent identifiers, claim identity verification, or overwrite a person’s chosen presentation in the name of consistency.

The hidden identity problem beneath a fluent translation

A person, place, organisation, work, product, species, law or dataset can have many names. A translator receives strings, but the reader needs referents: which individual, record or object the words denote. This distinction explains why a name can be correctly spelled and still be factually wrong. A translated article may describe the right event but attach it to the wrong account, city, product generation or historical individual. Source-to-target word similarity alone cannot prevent that failure.

Entity identity is not a claim that every identity must carry an official number. Many legitimate human names and cultural names have no single universal register. A robust system instead records the best available evidence for the reference, its certainty, the current preferred label and any context-specific variants. Where a stable public identifier exists, it can help disambiguation; where none exists, leave the matching status unresolved rather than constructing a fictional authoritative key.

The central workflow is recognition, resolution, representation and verification. Recognition finds a mention that refers to a specific entity. Resolution decides which entity the mention denotes and what evidence supports that decision. Representation chooses a suitable visible name in the target language or script. Verification confirms that the chosen output still points to the same entity. These are separate activities, and a team can succeed at one while failing another.

Thirty-six mechanisms that preserve names and referents

1. Entity versus label

The same person may have a native-script name and a preferred public Latin spelling. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Do not equate a visual string with a unique individual. The critical failure mode is to store the referent and its authorised labels separately. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A newsletter mentions an author by their publishing name while a source archive uses a different form. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

Diagnostic and transfer: show the target line to a reader who cannot see the source and ask them to locate the intended entity. Then test the same record in a citation, short caption, search result and formal database field. The identity anchor should remain stable even where the visible label changes with the format. Where disagreement arises, the reviewer should preserve both observed labels and attach a question to the source owner rather than silently overwriting the field.

2. Original-script preservation

A written form may carry distinctions lost when converted to another script. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Keep the original string available whenever authority and privacy permit. The critical failure mode is to treat transliterated output as a related representation rather than a replacement original. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A bibliographic catalogue retains a Cyrillic title alongside a Latin rendering. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

3. Transliteration

A mapping between writing systems is not necessarily a translation of meaning. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Record the chosen mapping scheme and its exceptions. The critical failure mode is to never assume that a generated Latin spelling is the person’s official preferred form. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A museum catalogue must distinguish its transliterated display from the accession record. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

4. Phonetic transcription

A pronunciation aid describes sound and is not always a reversible spelling system. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Choose appropriate pronunciation notation only when that is the reader’s job. The critical failure mode is to do not turn an approximate sound guide into a civil identity field. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A podcast pronunciation guide accompanies the written name without replacing it. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

5. Established exonym

A place can have a widely conventional name in another language. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Use a well-supported target usage when context and editorial policy justify it. The critical failure mode is to keep the source and target place linked to one verified geographic entity. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A map article mentions a city in a name customary to the target audience. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

6. Endonym and local form

The local name is important evidence of naming and historical context. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve the local form when it is relevant to identity and agency. The critical failure mode is to avoid presenting one community’s label as universally preferred by all speakers. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A cultural-history text discusses contested naming practices. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

7. Multiple people one spelling

Exact display-name equality does not establish person identity. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Require additional evidence such as a role date affiliation or unique identifier. The critical failure mode is to never merge two biographical profiles by string matching alone. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: Two researchers share the same romanised full name. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

8. One person multiple spellings

Legitimate variants may result from conventions publication history or personal choice. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Record known aliases with type and evidence. The critical failure mode is to do not automatically count different spellings as different people. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An author appears in journals under two Latin spellings. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

9. Preferred self-identification

People may choose how they are addressed and how names appear publicly. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Capture preference and context without assuming it changes legal records. The critical failure mode is to respect consent and avoid involuntary disclosure of past or private names. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An interviewee requests a publication name distinct from a registration name. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

10. Legal name versus display name

Official documents and ordinary editorial references may require different fields. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Use exact source authority requirements for official purposes. The critical failure mode is to do not substitute a nickname for a registration field or vice versa. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A scholarship announcement needs a public name but the payment process needs verified records. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

11. Mononyms

Some people have one name without a separate family name. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Design content and data fields that do not invent a missing surname. The critical failure mode is to do not copy a title or placeholder into surname storage. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A conference registration system must issue a badge to a one-name speaker. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

12. Name-part order

Given-first and surname-first patterns are not universal. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Store name components only when known and format according to context. The critical failure mode is to do not infer surname solely from leftmost or rightmost position. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A translated programme reorders a guest’s name for display. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

13. Compound surnames

A multiword surname may be one logical unit. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve the person’s own field boundaries. The critical failure mode is to do not split a compound into presumed middle names. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A certificate layout must keep a two-word family name intact. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

14. Multiple surnames

More than one family-name field may be culturally appropriate. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Follow documented personal naming structure when formats permit it. The critical failure mode is to do not drop the second surname to fit an assumed two-field model. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A researcher credit needs both surnames as used in publications. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

15. Patronymics and matronymics

An inherited parental-name element is not always a family surname. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Capture the actual linguistic relationship without imposing another system. The critical failure mode is to do not generate a false hereditary surname. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An oral history includes a person whose name indicates parentage. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

16. Surname particles and prefixes

Particles such as de van or von vary by name and context. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Follow the recorded segmentation and preferred casing. The critical failure mode is to do not strip or capitalize particles blindly. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A citation list must index an artist without damaging the display name. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

17. Honorifics and titles

A courtesy title office or degree is not automatically part of a person’s name. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Separate name data from address style and role. The critical failure mode is to do not invent gender marital status or credentials from appearance. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A letter introduces a speaker with a professional title. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

18. Diacritics

Accents and other marks may distinguish names and identities. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve authoritative spelling and Unicode fidelity. The critical failure mode is to do not silently fold characters in visible publication fields. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A registration ticket reproduces a name with an accented vowel. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

19. Unicode normalization

Canonically equivalent strings can have different code-point sequences. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Apply well-understood normalization at appropriate system boundaries. The critical failure mode is to do not confuse canonical equivalence with proof of identity. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A document ingester compares composed and decomposed accented strings. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

20. Compatibility mappings

Compatibility normalization may transform text more aggressively than NFC. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Document every matching transformation and preserve originals. The critical failure mode is to do not alter visible historical names using search-only folding. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An index supports tolerant lookup without changing displayed labels. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

21. Confusable characters

Some different Unicode characters look similar in some fonts. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Use script-aware review and security mechanisms in high-risk contexts. The critical failure mode is to do not assume visual similarity validates the same account. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An invoice contains a sender name with a misleading lookalike character. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

22. Mixed-script names

One label can contain letters from more than one writing system. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve the intended script and assess suspicious mixtures carefully. The critical failure mode is to do not strip meaningful characters merely because scripts differ. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A multilingual organisation uses a legitimate bilingual logo name. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

23. Bidirectional text

Arabic or Hebrew labels embedded in another direction can reorder nearby punctuation. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Use direction isolation in supported markup and test actual rendering. The critical failure mode is to do not trust a visually correct editor screenshot as proof the stored order is correct. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A bilingual citation displays a right-to-left title beside an identifier. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

24. Language tags

A language or script subtag describes text processing context not a person’s identity. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Use BCP 47 labels only where they describe actual content or presentation. The critical failure mode is to do not use a locale code as a substitute for citizenship or origin. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An accessible catalogue marks a script variant without changing the entity key. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

25. Locale-sensitive display

Name order initials spacing and informal forms can depend on locale and context. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Use relevant CLDR person-name patterns where applicable. The critical failure mode is to do not assume a formatter can parse an arbitrary name or verify its authenticity. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An event site produces a greeting and a formal programme credit. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

26. Sorting versus display

Collation keys and citation ordering need not match visible name wording. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Keep search sort and display functions separated. The critical failure mode is to do not print a sort key as the person’s chosen public name. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A library indexes a multiword surname but displays the authorised form. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

27. Aliases and former names

A person or organisation may have documented earlier names. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Record alias type validity and appropriate disclosure policy. The critical failure mode is to do not expose a sensitive former name without a lawful reason. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An archive links an old corporate name to the present entity. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

28. Pseudonyms and pen names

A creator’s public identity can legitimately differ from civil identity. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Retain the publication name and any authorised bibliographic relationship. The critical failure mode is to do not reveal a private identity or treat one as an error. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A novel credit uses a pen name consistently in translated editions. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

29. Organisation renaming

A name change may not create a new organisation and a merger may create one. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Check continuity and official registry evidence before linking records. The critical failure mode is to do not collapse a successor with a distinct predecessor. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A company history distinguishes a legal successor from a renamed firm. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

30. Product and brand variants

Brand names and model codes can combine protected words and descriptive text. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve invariant codes and verify authorised target labels. The critical failure mode is to do not translate a model code into a different product variant. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A device warning names a specific hardware revision. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

31. Scientific nomenclature

Scientific names and common names serve different taxonomic functions. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve authority and rank where needed and verify current accepted taxon. The critical failure mode is to do not replace a species name with a broad common-name category. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A conservation notice identifies a specific species and its synonym. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

32. Artwork and work titles

A work can have an original title an established translated title and edition-specific labels. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Link title variants to the same verified work or edition. The critical failure mode is to do not merge two different works because their translated titles match. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A book review cites an original edition and a translated reissue. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

33. Place-name ambiguity

Multiple geographic entities can share the same name. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Resolve by reliable context and geographic reference rather than spelling alone. The critical failure mode is to do not choose the most famous place merely because it appears first in search. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A travel article mentions a town sharing a name with a major city. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

34. Coordinates and named places

A place label and a point or polygon represent different geographic information. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Check coordinate system scale and boundary semantics separately. The critical failure mode is to do not treat one coordinate as an exhaustive definition of an area. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An environmental map labels a protected zone. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

35. Historical names and periods

A place or institution may have different names during different eras. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Preserve period-specific names where history and evidence require it. The critical failure mode is to do not retroactively attribute current names to every historical voice. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: A museum timeline quotes a document using a former place name. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

36. Identifiers and authority records

Stable identifiers can help connect labels across languages. The translation task is to preserve the relation between the real-world referent and the string that describes it. A competent reviewer asks whether the source is naming the entity, describing a label, documenting a historical form or issuing a binding identifier. That question comes before the choice of transliteration, borrowing, translation or localised formatting. Otherwise the target may be readable while directing the reader to a different person, item or record.

Preservation rule: Check the issuing authority scope current status and collision risk. The critical failure mode is to do not invent a universal ID or treat all registries as equivalent. Reversibility deserves attention: a transliteration may preserve enough clues for human recognition while still losing distinctions needed to recover the exact source writing. Keep the supplied original alongside the approved target form when the document’s purpose allows, and record whether any mapping was authorised by the named person, an issuing institution or an editorial policy.

Worked scenario: An archive references an authority record for a contributor. At intake, write down the original mention exactly, including script, punctuation, spaces, diacritics and the source document’s date. Decide which of those features identify the referent and which are presentational. An acceptable target can use an established equivalent, but it must not introduce a false biography, credential, organisation, region or version. If two candidate entities remain plausible, explicitly mark the identity unresolved instead of choosing the most convenient one.

The twenty-two-pass identity-safe translation method

Pass 1: Define the reader’s job

Identify whether a user must address a person locate a place cite a work buy a product or match a record. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: An error matters differently for a museum label and a beneficiary-payment field. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Transfer check: take this decision into an index, an on-screen label, a spoken pronunciation guide and a document citation. The visible presentation may differ by format, but each output should point back to the same evidence record. A change that breaks this correspondence is an identity error even if the phrase translates well. Record how it was repaired so parallel translations can be updated consistently.

Pass 2: Inventory named entities

Mark every person organisation place work product dataset species institution and identifier. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Do not stop at capital letters because scripts and naming conventions differ. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 3: Preserve original strings

Store the exact source label including punctuation diacritics script and whitespace. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Do not destroy originals through irreversible cleanup. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 4: Establish source provenance

Record the document version author date and authority for the label. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: A copied name is not necessarily a personally approved name. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 5: Assign provisional entity keys

Use a stable project-local identity reference pending authoritative resolution. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Do not present internal keys as public civil identifiers. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 6: Check ambiguous candidates

Gather contextual role time affiliation field and relationships. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Identical names alone do not justify merging. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 7: Select trusted authority

Identify appropriate source records for bibliographic place corporate or personal usage. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: No single registry governs every type of name. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 8: Ask for preference and consent

Where possible preserve the named person’s chosen public form. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Do not turn private aliases into public content by convenience. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 9: Classify each name segment

Determine whether it is a name part title affix suffix model code or descriptor. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Avoid arbitrary surname parsing or translating identifiers. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 10: Distinguish transformation types

Choose between established translation exonym transliteration transcription borrowing and unchanged form. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: A pronunciation guide is not a legal spelling. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 11: Choose language and script metadata

Annotate actual source and target language or script when systems require it. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: A BCP 47 language tag is not an identity record. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 12: Check locale display needs

Decide order formality initials and name length for the output channel. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Formatting does not authorise changing who is named. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 13: Maintain approved variants

Record original script official transliteration editorial equivalent alias and historical name with types. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Unapproved variants cannot be silently promoted. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 14: Protect technical invariants

Keep identifiers codes ISBNs specimen IDs and registry numbers exact where relevant. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: A single character change may select another record. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 15: Handle bidirectional strings

Test right-to-left inline names next to dates numbers and punctuation. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Visible order and stored order are distinct. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 16: Review Unicode handling

Apply safe canonical normalization deliberately while preserving original presentation. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Compatibility folding and confusable handling need separate risk checks. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 17: Validate target referents

For each translated mention follow it back to the intended identity and original evidence. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Do not rely on surface string similarity. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 18: Check cross-document consistency

Confirm that captions tables figures headings footnotes and citations use related authorised names. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: A document may be internally fluent yet inconsistent. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 19: Test search and discoverability

Ensure approved aliases can retrieve the original entity without corrupting display forms. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Search matching is not identical to publication style. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 20: Review privacy and safety

Consider whether alternate names sensitive histories or identifying attributes may be disclosed. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Translation should not create new privacy exposure. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 21: Run independent QA

Give reviewers target-only identity lookup tasks and ambiguous near-match cases. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: A correct-looking name is not proof of correct mapping. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Pass 22: Log decisions and revisions

Document mappings evidence unresolved cases owner and change control. A practical deliverable from this step is not a vague tick mark. It is a record that shows what was received, what was inferred, what authoritative evidence was consulted and which version of the target text used the decision. A source-to-target entity map is stronger than a sheet of preferred spellings because it connects the names to referents, not merely to other strings.

Guardrail: Future teams must know why a spelling was chosen. For a high-impact identity, have a second reviewer attempt to distinguish plausible alternatives. They should be able to answer why this target label denotes this entity and not another one with similar spelling. If the answer depends on a private assumption, an undocumented search result or a guessed name component, mark the issue unresolved and request clarification.

Thirty-two failure modes: how fluent translations attach words to the wrong entities

This failure register is separate from general spelling advice. Each defect describes a change in what or whom the translated sentence identifies. Severity depends on whether the wrong mapping could misattribute speech, redirect a payment, invalidate a citation, send someone to the wrong place or confuse a model with another model. In a harmless literary discussion the consequence may be interpretive; in a technical or legal record it can be operational.

Failure 1: Replaced by famous namesake

Observed defect: An uncommon person is replaced with a celebrity sharing the name. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Role and source context disambiguation. A search popularity ranking does not prove identity. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 2: Surname invented for mononym

Observed defect: A system inserts a placeholder surname and publishes it. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Keep a true single-name field and format properly. Validation convenience is not authority to create a second name. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 3: First and last name reversed

Observed defect: An unfamiliar given-first or family-first structure is guessed. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Check native structure or user-supplied segmentation. Name order and component role are different issues. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 4: Compound family name split

Observed defect: One surname is separated into surname and middle-name fields. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Preserve the documented whole surname. Whitespace is not a universal component boundary. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 5: Second surname omitted

Observed defect: A multi-surname creator appears under a truncated credit. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Restore the authorised full attribution. Citation brevity must not erase identity. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 6: Patronymic called surname

Observed defect: A parental-name element becomes a permanent inherited surname. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Keep the relation as documented. Family-structure assumptions distort the name. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 7: Honorific translated as name

Observed defect: Dr or an office title is stored as a given name. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Separate title and personal-name fields. Credentials and names carry different facts. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 8: Title mistaken for office

Observed defect: A translated courtesy title suggests a role not held. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Verify professional and institutional standing. Do not confer authority with style. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 9: Diacritic silently lost

Observed defect: An accented name is collapsed in public output. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Preserve the source and approved display spelling. Search folding and display are different purposes. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 10: Compatibility fold destroys text

Observed defect: A search-normalised form overwrites the original record. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Store original separately from matching key. NFKC decisions need careful contextual review. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 11: Confusable account mistaken

Observed defect: Visually similar characters point to another identifier. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Apply script-aware validation and known-source comparison. A visual match is not a secure identity match. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 12: Wrong-script assumption

Observed defect: A language tag is mistaken for the person’s preferred script. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Check actual text script and context. Locale does not establish personal identity. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 13: Incorrect romanisation scheme

Observed defect: A name is converted with a system inconsistent with source authority. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Document scheme and prefer authorised form. Automatic transformations can be nonreversible. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 14: Pronunciation spelling promoted

Observed defect: A phonetic aid becomes the official written name. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Retain phonetic guide as a separate field. Approximate pronunciation does not equal legal spelling. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 15: Unapproved endonym substitution

Observed defect: An editor replaces a historic exonym without checking context. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Preserve period and intended audience distinctions. Historical quotations have their own wording. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 16: Wrong place selected

Observed defect: A familiar city is chosen for a smaller namesake town. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Use source coordinates jurisdiction or contextual evidence. Name frequency does not resolve location. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 17: Boundary mistaken for point

Observed defect: One geocoded pin is treated as the complete geographic feature. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Keep geometry type and reference. A pin may represent a zone only approximately. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 18: Renamed corporation merged

Observed defect: Two entities connected by ownership are treated as one legal person. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Inspect registration and continuity details. A merger is not always a simple rename. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 19: Former name disclosed

Observed defect: An individual’s sensitive previous name becomes searchable public copy. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Follow consent privacy and publication purpose. Retrievability is not permission to publish. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 20: Nickname assumed official

Observed defect: A friendly name appears on a contract or certificate. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Use the correct verified document form. An address preference and official field serve different needs. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 21: Pseudonym unmasked

Observed defect: A translator links a public pen name to a private civil identity. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Respect authorised public identity only. Bibliographic cross-reference may be restricted. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 22: Brand descriptor overtranslated

Observed defect: A brand mark is replaced with a generic translated phrase. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Preserve brand identity and translate only the surrounding descriptor. Brand recognition is not mere vocabulary. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 23: Model number mutated

Observed defect: A translated product code now denotes a different configuration. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Protect alphanumeric invariant strings. A single suffix may determine hardware compatibility. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 24: Taxon synonym misapplied

Observed defect: A common name is mapped to a different biological species. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Confirm taxonomic authority and context. Lexical similarity is not taxonomy. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 25: Book title conflated

Observed defect: Two works with similar translated titles are treated as one. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Check author edition publication and work identifiers. A title is not a unique catalogue key. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 26: Person conflated with organisation

Observed defect: A founder’s name is mistaken for the organisation they founded. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Track entity type and relationship. Name overlap can represent association not identity. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 27: Role conflated with person

Observed defect: A job title is treated as a permanent name. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Preserve incumbent and role separately. An office can have successive holders. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 28: Citation initials overcompressed

Observed defect: Shortened initials make two authors indistinguishable. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Use disambiguating source metadata. Formatting must preserve attribution. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 29: Alias treated as new entity

Observed defect: A spelling variant causes duplicate profiles. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Attach verified variants to one reference. Separate spellings may share a referent. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 30: Distinct entities collapsed

Observed defect: Two accounts with similar variants are automatically joined. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Require independent identity evidence. Fuzzy matching is not confirmation. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 31: Incorrect bidi rendering

Observed defect: A mixed-script identifier displays punctuation in misleading order. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Use direction isolation and verify logical storage order. Visual inspection alone is insufficient. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Failure 32: Identity overwritten in update

Observed defect: A later alias update destroys the original authority trail. The resulting target may be perfectly grammatical and still refer to the wrong individual, edition, corporate entity or location. A reviewer should first compare the intended referent with what a reader can actually locate from the translated wording. Where the target makes a different entity appear responsible for a statement, the failure is semantic and documentary, not just typographical.

Repair: Version mappings and retain evidence. Translation maintenance is not record erasure. When two plausible candidates remain, mark the relation unresolved in the entity ledger and escalate rather than silently guessing. Verify the corrected target in both the complete paragraph and an isolated search result, citation or system label. A valid correction must survive the compact context where users may see the name without its explanatory text.

Worked cases: sixty identity-preserving translation diagnostics

The following cases are fictional and schematic. Each one uses a source situation, an unsafe move and a defensible action. They are semantic demonstrations, not invented translations into unfamiliar languages, nor evidence that any particular person, company or product exists. A production translator should check official spellings, actual source-script names and local publication rules before applying the method to real people or high-stakes records.

Diagnostic workshop 1: Personal introductions

The personal introductions examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 1: Personal introductions — A visiting researcher submits one preferred public

Source situation: A visiting researcher submits one preferred public name and a longer administrative form. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: An editor forces every credit into the administrative form. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The conference programme uses the submitted public name while secure registration retains the verified full form. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Who supplied and authorised each form? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: preserve separate public and restricted fields; do not expose a private component. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Personal agency and auditability must coexist. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 2: Personal introductions — A speaker has a one-word name that the invitation

Source situation: A speaker has a one-word name that the invitation platform will not accept. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator invents a family name from the first letters of the speaker’s organisation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The invitation identifies the one-name speaker accurately and refers the form-design constraint to an administrator. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Can the system represent a one-part name? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not fake a second personal name as data. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Software constraints are not linguistic facts. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 3: Personal introductions — A programme contains a two-word given name and a c

Source situation: A programme contains a two-word given name and a compound family name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated credit splits both at spaces and reverses the wrong components. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The reviewer preserves the supplied name-part boundaries and chooses the agreed display order. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Are the component boundaries recorded by the person? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: treat raw whitespace as presentation rather than proof of name structure. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Incorrect parsing can create an entirely different citation. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 2: Historical biographies

The historical biographies examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 4: Historical biographies — A source calls an artist by a name used earlier in

Source situation: A source calls an artist by a name used earlier in their career. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated text replaces every historical occurrence with the modern name. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The article explains the historical name once and uses context-appropriate references thereafter. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which form was current during the event? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep original quoted wording and annotate relevant later names. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Historical accuracy sometimes requires more than one label. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 5: Historical biographies — Two individuals share a romanised name in the same

Source situation: Two individuals share a romanised name in the same period. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator combines their achievements into a single biography. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The biographies remain distinct using evidence from role, dates, works and archives. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is there independent confirmation the records describe one person? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: maintain separate provisional entity keys until resolved. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A plausible surname match is not proof. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 6: Historical biographies — A quotation contains an old institution name that

Source situation: A quotation contains an old institution name that later changed. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation retroactively puts the modern institution’s name inside the quotation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The original quotation remains intact and a nearby explanation identifies any supported successor relationship. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the later institution have continuous identity? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: separate source voice from editor explanation. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Correction of readers’ context must not rewrite the past. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 3: Scientific authorship

The scientific authorship examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 7: Scientific authorship — A researcher publishes under two legitimate Latin-

Source situation: A researcher publishes under two legitimate Latin-script spellings. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A bibliography treats both as distinct authors solely because the strings differ. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The editorial ledger links the spellings only after verifiable author evidence. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Do the publications themselves provide an identifier or consistent affiliation? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: preserve both citation strings with documented relation. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Search recall and correct attribution have different functions. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 8: Scientific authorship — Two authors publish in different disciplines under

Source situation: Two authors publish in different disciplines under the same initials. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation expands both initials to one guessed person. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Each record keeps its verified author form without invented expansion. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What original metadata identifies the authors? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: query a reliable catalogue instead of expanding automatically. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: An abbreviation can increase ambiguity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 9: Scientific authorship — An article title has a published translated versio

Source situation: An article title has a published translated version and an independent work with similar words. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A summary cites the wrong work because its title matches the translation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The record checks author, edition, publication data and stable work identifier. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which edition or work is actually cited? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep work identity distinct from title rendering. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Translation creates additional labels, not permission to merge works. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 4: Library catalogues

The library catalogues examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 10: Library catalogues — A catalogue stores an original-script title and a

Source situation: A catalogue stores an original-script title and a Latin-script transliteration. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: An import routine overwrites the source field with the Latin output. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Both original title and labelled transliteration are retained as separate related data. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the mapping reversible and which system produced it? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep source characters for authority comparison. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Recoverability matters for bibliographic evidence. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 11: Library catalogues — An index sorts names by a culturally appropriate f

Source situation: An index sorts names by a culturally appropriate field. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation prints the sort key as the person’s displayed name. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Sorting metadata remains independent of the approved visible label. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the reader-facing form match the authorised credit? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: test sorting and display separately. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: An index key is not a public self-identification. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 12: Library catalogues — A pseudonymous writer’s works are connected to an

Source situation: A pseudonymous writer’s works are connected to an authority record. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator reveals a restricted identity in a public annotation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The publication uses only names and relations permitted for that context. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the linked identity public and authorised? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: check source sensitivity before hyperlinking. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Cross-reference capability is not consent. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 5: Contact management

The contact management examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 13: Contact management — A person supplies their name with accents in a cus

Source situation: A person supplies their name with accents in a customer profile. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A cleanup script strips the accents in every exported greeting. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Display and correspondence retain the supplied form; a search key may use documented tolerant matching. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the exported name the version the person chose? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: separate matching normalisation from visible text. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Names deserve the same care as the surrounding message. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 14: Contact management — A profile has a formal name and preferred form of

Source situation: A profile has a formal name and preferred form of address. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator assumes the first given-name token is the preferred salutation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The user is asked how they prefer to be addressed and the answer is stored separately. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What address form has the person actually chosen? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not infer intimacy from database structure. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Greeting conventions are not mechanical slicing. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 15: Contact management — A contact record has an unknown surname field

Source situation: A contact record has an unknown surname field. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: An export inserts a placeholder that becomes searchable as a real surname. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The field remains unknown and the UI uses a respectful available display label. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the person genuinely have a family name field? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: avoid fake data to satisfy validation. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Missing metadata is not an invitation to fabricate. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 6: News publishing

The news publishing examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 16: News publishing — A public statement refers to a company using a his

Source situation: A public statement refers to a company using a historic trading name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation treats the trading name as an unrelated firm. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The article documents verified continuity or labels the relation as unresolved. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Was there a legal entity change or only a brand change? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: check official registries and original reporting. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Corporate naming needs dates and authority. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 17: News publishing — A named office holder speaks in an official capaci

Source situation: A named office holder speaks in an official capacity. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated account attaches the statement to the office itself forever. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The statement is attributed to the correct person and their role at the time. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Who actually made or authorised the statement? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: distinguish office, incumbent and organisation. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Roles change even when office names remain. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 18: News publishing — A place appears under an established target-langua

Source situation: A place appears under an established target-language exonym. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator chooses a different nearby settlement with a similar name. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The approved place name is chosen only after matching geography and documentary context. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which place is referenced by source location evidence? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: use verified geographic identity rather than surface familiarity. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A conventional name is not a geocoding result. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 7: Museum exhibitions

The museum exhibitions examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 19: Museum exhibitions — An artefact has an original title, a translated di

Source situation: An artefact has an original title, a translated display title and an accession number. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A translated panel mistakenly treats a second artefact with the same translated title as the first. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The accession record anchors identity and each display title is linked to it. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which accession object does each label describe? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: protect the identifier across captions and maps. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Titles may be nonunique while accession IDs have scoped uniqueness. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 20: Museum exhibitions — A collection uses a traditional place name no long

Source situation: A collection uses a traditional place name no longer current. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation updates the historical wording without explanatory context. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The panel retains historical usage and explains modern names where relevant. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What period and speaker does the label describe? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: avoid projecting today’s language into period testimony. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Display names can be historically situated. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 21: Museum exhibitions — An artist’s work is popularly known under a shorte

Source situation: An artist’s work is popularly known under a shortened title. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A summary treats the short title as a new work. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The short title is recorded as a variant of the verified work. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What catalogue evidence connects the labels? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: separate alias from edition and work identity. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Shortening changes search text, not the object. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 8: Product support

The product support examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 22: Product support — A device line has two revisions separated by one m

Source situation: A device line has two revisions separated by one model suffix. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated manual drops the suffix and gives steps for the wrong hardware. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The exact model identifier is protected while surrounding descriptions are translated. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which hardware revision does the instruction apply to? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: use invariant code and approved product name. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Model tokens can be safety-critical identifiers. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 23: Product support — A brand name looks like an everyday source-languag

Source situation: A brand name looks like an everyday source-language word. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator renders it as an ordinary noun in a warranty message. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The brand is preserved using the approved mark, with generic terms translated separately. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the term a registered or authorised product name? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: check product nomenclature rather than dictionary sense. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Proper names and descriptive vocabulary have different jobs. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 24: Product support — A support database contains several regional marke

Source situation: A support database contains several regional marketing names for one SKU. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation assumes all similar labels are identical without checking SKU mapping. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Only approved marketing names are linked to the same verified product record. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the SKU or variant actually identical? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep sale region and generation as separate fields. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: One product can carry many labels, but not every label is the same product. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 9: Geographic guides

The geographic guides examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 25: Geographic guides — Two small towns share a name but appear in differe

Source situation: Two small towns share a name but appear in different districts. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator picks the most frequently returned search result. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The target name is matched to the specific source district and reliable geographic reference. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which district and coordinates are supported? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: leave unresolved if source context is insufficient. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Search popularity cannot choose geography. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 26: Geographic guides — A route mentions a protected landscape represented

Source situation: A route mentions a protected landscape represented as a map pin. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated guidance treats the pin as an exact legal boundary. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The route retains the place identity while describing the map pin as approximate. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What spatial representation does the source actually provide? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep entity identity separate from geometry. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: One marker is not an entire boundary. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 27: Geographic guides — A public notice uses a disputed or multilingual pl

Source situation: A public notice uses a disputed or multilingual place name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation declares one historical label universally authoritative. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The publication follows the relevant current naming policy and carefully notes contested context. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which community and authority define this usage? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: avoid anachronism and unsupported neutrality claims. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Naming can carry social and historical meaning. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 10: Software interfaces

The software interfaces examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 28: Software interfaces — A user account name contains letters from multiple

Source situation: A user account name contains letters from multiple scripts. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator changes lookalike characters to make it easier to read. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The verified identifier stays intact and suspicious variants are flagged for security review. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the change a known safe canonical mapping? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not silently alter credentials or account IDs. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Visual similarity is not identity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 29: Software interfaces — A screen reader displays an Arabic name beside a L

Source situation: A screen reader displays an Arabic name beside a Latin serial number. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated UI leaves punctuation reordered in visual output. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Direction isolation is applied in supported markup and actual rendering is tested. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the stored character sequence match the intended string? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: test both visual and logical order. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Bidi failures can make names and numbers appear attached incorrectly. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 30: Software interfaces — A locale field says a document is in Serbian Latin

Source situation: A locale field says a document is in Serbian Latin script. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation interprets the tag as proof the named owner has a particular nationality. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The tag is used only to control language-script rendering. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What does the metadata field actually describe? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not turn language into personal origin. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Language labels and identity are different ontologies. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 11: Clinical records

The clinical records examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 31: Clinical records — A patient name arrives with official document spel

Source situation: A patient name arrives with official document spelling and a different everyday display form. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated chart replaces the legally verified record with the informal form. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The appropriate verified identity field remains exact and a separate permitted display form is used. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which field is required for clinical identification? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: use approved matching procedures and privacy controls. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Never treat a nickname as a clinical identifier. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 32: Clinical records — Two patients share a name and date format causes a

Source situation: Two patients share a name and date format causes a mix-up. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator claims the name uniquely identifies the record. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The message uses authorised record-specific identifiers without exposing them unnecessarily. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What second factor distinguishes the records? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: human safety requires reliable patient identification controls. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Names are not sufficient identifiers in every workflow. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 33: Clinical records — A lab specimen has a human-readable nickname and a

Source situation: A lab specimen has a human-readable nickname and an exact barcode. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation renders the nickname while mistyping the barcode. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The barcode is protected as the operational invariant and the display name reviewed separately. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the specimen label map to the same laboratory record? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: cross-check all machine-readable identifiers. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A pleasant label cannot repair wrong specimen identity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 12: Financial documents

The financial documents examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 34: Financial documents — A beneficiary record uses a legal name plus a pref

Source situation: A beneficiary record uses a legal name plus a preferred correspondence name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated payment instruction substitutes the greeting name for verified payee identity. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The payment field remains aligned with its authorised source record. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which name does the receiving institution require? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not guess legal forms or payment requirements. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Incorrect identity fields can redirect or block transactions. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 35: Financial documents — Two companies with similar abbreviated names appea

Source situation: Two companies with similar abbreviated names appear in a report. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator expands both to the same guessed organisation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The source retains distinct entities until separate registry evidence resolves each. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What authoritative record supports each expansion? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: maintain ambiguity in draft rather than merge. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: An acronym may not be globally unique. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 36: Financial documents — An organisation has a historical name and a new tr

Source situation: An organisation has a historical name and a new trading identity. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated chronology presents a later brand as if it existed in every earlier year. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Each transaction is labelled with the name valid in its original documentary context. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: When did the documented name change? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep date-scoped mappings. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Revision should preserve history instead of flattening it. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 13: Literary translation

The literary translation examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 37: Literary translation — A fictional character uses an alias later revealed

Source situation: A fictional character uses an alias later revealed to refer to another named character. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator reveals the identity prematurely through an explanatory target name. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The target preserves the narrator’s knowledge at each point of the story. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: When does the reader learn the alias relationship? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not resolve plot information before the source does. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Identity preservation includes deliberate narrative ambiguity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 38: Literary translation — Two characters have deliberately similar sounding

Source situation: Two characters have deliberately similar sounding names. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator makes their names identical to simplify pronunciation. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The target preserves a distinguishable pairing consistent with the work’s authorial constraints. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the difference central to the plot? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: maintain a character-name concordance. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Names can be narrative devices, not mere labels. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 39: Literary translation — A work uses meaningful descriptive names that invi

Source situation: A work uses meaningful descriptive names that invite adaptation. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator mechanically translates every name without examining intended effect. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Decisions distinguish semantic adaptation from stable identity and remain consistent across the work. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which names are semantically active and which should be retained? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: document creative choices and reader consequences. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Creative equivalence still needs identity continuity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 14: Scientific species

The scientific species examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 40: Scientific species — A common bird name is used for several related spe

Source situation: A common bird name is used for several related species in different languages. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated report chooses a familiar but incorrect species. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The scientific taxon identifier or accepted name anchors the match. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What taxonomic concept is supported by the observation? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: avoid one-to-many common name substitution. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Species identity is not solved by dictionary lookup. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 41: Scientific species — An older paper uses a historical taxonomic synonym

Source situation: An older paper uses a historical taxonomic synonym. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation silently upgrades the old name without explaining the change. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The historical designation is preserved and any current synonym is linked with dated authority. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Did the taxonomic concept itself change? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: track name and concept separately. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Synonymy can reflect changes in scientific understanding. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 42: Scientific species — A specimen label includes a precise accession numb

Source situation: A specimen label includes a precise accession number. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation copies a visually similar but distinct code point. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The accession string is checked character by character against source authority. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the record resolve to the correct specimen? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: keep code and human-readable name distinct. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A subtle typographic change can create a false reference. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 15: Data catalogues

The data catalogues examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 43: Data catalogues — A research dataset is known by a short label and a

Source situation: A research dataset is known by a short label and a persistent identifier. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated abstract links the label to a different similarly named dataset. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The persistent identifier and version are used to verify the intended source. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Does the citation resolve to the original dataset version? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: protect identifier and version separately. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A title alone is not a robust data citation. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 44: Data catalogues — A list contains two unrelated projects with simila

Source situation: A list contains two unrelated projects with similar acronyms. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A machine-assisted translation expands both to the more familiar project name. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Each project retains its source-specific acronym until verified expansion. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the expansion supported by local context? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: avoid global acronym replacement. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Contextual disambiguation is essential. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 45: Data catalogues — An archived file has a renamed successor but diffe

Source situation: An archived file has a renamed successor but different contents. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation treats the successor as the same immutable file. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The record distinguishes related but separately versioned data artefacts. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the checksum or version identical? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: do not use similar titles as data identity proof. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Version and entity relationships must be explicit. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 16: Social histories

The social histories examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 46: Social histories — An interviewee uses a self-chosen name in their pu

Source situation: An interviewee uses a self-chosen name in their public account. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated article substitutes an older registry name found in background notes. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The published piece respects the agreed public name and confidentiality. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Was disclosure of the registry name authorised? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: separate source access from publication rights. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Identity preservation has ethical boundaries. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 47: Social histories — A community organisation has a local-language name

Source situation: A community organisation has a local-language name and an international acronym. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation converts the acronym to a guessed expansion. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The approved acronym and authorised full names are recorded together. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Who issued the accepted name? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: avoid inventing institutional expansions. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Community self-naming matters. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 48: Social histories — An oral history uses several nicknames for one per

Source situation: An oral history uses several nicknames for one person. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation treats each nickname as a new speaker. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Only verified aliases are linked under the same entity while uncertainty remains marked. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Did the narrator explicitly connect those names? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: maintain a source-specific alias map. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Oral records depend on social context, not database guesswork. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 17: Editorial newsrooms

The editorial newsrooms examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 49: Editorial newsrooms — A photo caption names someone using the wrong sour

Source situation: A photo caption names someone using the wrong source-script variant. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The newsroom trusts a Latin spelling supplied by an unrelated image search. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The caption is checked against original assignment or verified identity data. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Who is actually pictured according to trusted source records? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: reject resemblance-only identification. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A photograph title is not proof of pictured identity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 50: Editorial newsrooms — A headline uses a short organisational name that s

Source situation: A headline uses a short organisational name that several groups share. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated headline drops the distinguishing context. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The headline adds sufficient role or context to prevent mistaken attribution. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Can an unfamiliar reader identify which organisation is meant? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: check isolated-title interpretation. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Short-form writing should not create misleading identity. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 51: Editorial newsrooms — A corrected article changes a person’s displayed n

Source situation: A corrected article changes a person’s displayed name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation updates the headline but not image alt text or embedded cards. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: All language and media surfaces update from a verified mapping record. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Where does the old name remain live? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: audit derivative artefacts. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Publication systems multiply identity references. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 18: Education materials

The education materials examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 52: Education materials — A biography lists an artist whose surname is also

Source situation: A biography lists an artist whose surname is also a place name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: A translation treats the surname as a geographical description. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The name is kept as a person’s proper name and its referent checked. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the phrase a person or place mention? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: use grammatical and topical context. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Capitalisation alone is insufficient. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 53: Education materials — A lesson adapts a culturally specific form of addr

Source situation: A lesson adapts a culturally specific form of address. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator converts a respectful address into an unrelated credential. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The role and interpersonal relation remain accurate without inventing professional status. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: What does the original address form convey? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: explain conventions if required. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Register can carry identity claims. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 54: Education materials — An exam excerpt includes two names that differ by

Source situation: An exam excerpt includes two names that differ by a small spelling feature. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation regularises them into one form. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: Both distinct character names remain distinguishable as source intended. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Are they different referents in the passage? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: preserve contrast, not just style. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Reader comprehension depends on identifying agents. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Diagnostic workshop 19: Transit information

The transit information examples show why an identity check must use the intended reader’s task. A display label may be adequate to recognise a person in prose, yet inadequate to identify a medical record or machine component. The following cases deliberately distinguish a written name from the referent and from the evidence used to choose a target-language representation.

Worked diagnostic 55: Transit information — Two stations share part of a name but serve differ

Source situation: Two stations share part of a name but serve different lines. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translated alert removes the qualifying station element. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The message retains the official station identity and line context. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which station is operationally intended? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: use authority-provided stop IDs where available. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A shortened location can misdirect a traveller. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 56: Transit information — A place label appears in an older service map and

Source situation: A place label appears in an older service map and a newer route map. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translation conflates historical and current station names. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The version and service-period context remain explicit. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Which map date governs this instruction? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: retain effective-date meaning. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: A same-place label can vary over time. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Worked diagnostic 57: Transit information — A route identifier resembles a translated place na

Source situation: A route identifier resembles a translated place name. A competent translator first identifies the referent and the source’s available evidence. The record may include a familiar label but lack a reliable identification field, or it may contain an exact code that looks unimportant in ordinary prose. Its function cannot be inferred from typography alone. The reviewer should mark which fields are identifiers, which are human-facing names, and which are mere descriptions.

Unsafe transformation: The translator changes the route code into a lexical equivalent. The mistake would make the target text point to an unsupported entity, describe an unsupported relationship or replace a source identity with a culturally convenient guess. Even when the translated sentence is fluent, a reader relying on it could connect evidence to the wrong person, work, location or object. The failure is not fixed simply by adding a generic explanation; the identity mapping itself must be examined.

Meaning-preserving solution: The route code is preserved and the descriptive name translated separately. This is a schematic editorial action rather than a claim that an automated translation has already verified any actual entity. In the working ledger, store original wording, any recognised script and language, exact identifier when provided, chosen display label, evidence for the connection and review status. If a trustworthy source cannot disambiguate alternatives, the output should accurately signal unresolved identity rather than invent an authoritative answer.

Diagnostic question: Is the code a machine-actionable field? An effective release test requires a reviewer to reproduce the mapping using the evidence recorded with the source, not merely approve that a name “looks right.” The operational safeguard is: protect operational identifiers. A second reader should also test how the same label behaves alone in search results, captions or forms where the surrounding paragraph is absent.

Transfer insight: Routes require both language and exact reference. Reformat the example as a headline, glossary entry, link label and index record; then ask whether the referent remains stable under those changes. If the visible string changes appropriately by language but the identity anchor stays constant, the translation has moved meaning rather than merely characters. If it acquires a different identity, the translation needs rework before publication.

Transfer exercises: make identity survive every output channel

The cases above train recognition of mistakes, but lasting skill requires transferring one correct entity mapping into several formats. A long article, a machine-readable catalogue record, an invitation, a subtitle and a compact button label all have different display needs. The project should not force identical typography everywhere; it should enforce identical referents, appropriate permission and traceable reasons for each variation.

Transfer task 1: Original-script retention

Setup: A translated history article needs a romanised display name but a quotation in the original script. The authoring job is to publish the source string and authorised transliteration with a clear explanation of which form is which. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check readers can use the original record to find the person without treating the Latin rendering as an official change. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 2: Name order in an invitation

Setup: A speaker’s native format puts a family name before a given name. The authoring job is to use supplied name-part metadata and an appropriate display pattern instead of guessing from space order. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check whether the speaker approves the greeting and programme credit. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 3: Two namesakes

Setup: Two professionals share an identical public name. The authoring job is to add verified role and context without fabricating a middle name. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Ask whether the resulting target allows readers to distinguish both records. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 4: Former institution name

Setup: A historical source uses a predecessor name. The authoring job is to preserve the original period reference and explain only supported successor relations. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check that the translation does not claim the later organisation existed before it did. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 5: Books with similar titles

Setup: Two works have target-language titles that look alike. The authoring job is to use author edition and verified identifiers to distinguish citations. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Verify that outbound links lead to the intended work and edition. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 6: Specimen labels

Setup: A translated field report includes a long accession code. The authoring job is to lock the exact code and translate only the descriptive label. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Compare code points and the authoritative accession record. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 7: Geographic ambiguity

Setup: A place name matches several settlements. The authoring job is to use context to establish one place, or label the ambiguity unresolved. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Do not use popularity rank as a substitute for geographic evidence. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 8: Brand translation

Setup: A named product includes ordinary descriptive words. The authoring job is to separate brand and model identifiers from the surrounding translatable category. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check the translated sentence still points to the intended variant. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 9: Preferred public identity

Setup: An interviewee uses a public name distinct from a restricted record. The authoring job is to publish only the agreed public form and keep restricted mappings permissioned. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Test captions hyperlinks and metadata for accidental disclosure. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 10: Bidi markup

Setup: A right-to-left name appears next to a Latin ID. The authoring job is to use direction isolation and test punctuation and number placement. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Compare logical stored order against displayed output. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 11: Accent-preserving display

Setup: A user enters a name with a non-ASCII letter. The authoring job is to retain the original on visible pages while allowing purpose-specific tolerant search. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Never let a search key overwrite the user-entered name. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 12: Transcription versus transliteration

Setup: An editor wants readers to pronounce a source-script name. The authoring job is to provide a pronunciation aid separately from transliteration and official spelling. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Ask whether the pronunciation guide is approximate and labelled. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 13: Aliases in an index

Setup: An archive uses three names for a documented single contributor. The authoring job is to link proven variants to one record with variant types. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check that search finds aliases without corrupting the public display. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 14: Unverified alias

Setup: Two online profiles look like one person. The authoring job is to keep identity unresolved until independent evidence supports the mapping. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Do not merge personal data on appearance or shared interests. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 15: Taxonomic common name

Setup: A species has several common names and a scientific name. The authoring job is to use the applicable scientific identity and approved vernacular label. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Test whether another species could plausibly match the target common name. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 16: Author citation style

Setup: A style guide requests initials that produce collision. The authoring job is to preserve the source citation and include sufficient disambiguation. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check the reader can retrieve the correct cited author. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 17: Official document field

Setup: An administrative translation contains a verified legal name and a nickname. The authoring job is to keep the legally required field faithful and isolate informal display use. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Obtain qualified review where actual identity verification matters. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 18: Search results

Setup: A platform localises names but also retains native-script forms. The authoring job is to display suitable forms while indexing supported aliases separately. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Test false positives and false negatives with namesakes. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 19: Speech and captions

Setup: A narrator says a person’s name but subtitles use a different variant. The authoring job is to align both forms with the same verified individual. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check captions do not invent a different person or alter who spoke. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 20: Responsive interface

Setup: A mobile layout shortens a long institutional name. The authoring job is to use approved short label plus accessible full context. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check whether the shortened form collides with a different institution. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 21: Name correction

Setup: A public article changes an incorrect spelling after reader feedback. The authoring job is to verify the correction and update related surfaces without silently changing archived quotations. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Record effective date and correction reason. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 22: Historical dispute

Setup: Two communities use differing labels for a location. The authoring job is to respect the publication context and avoid unsupported claims of universal preference. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Preserve any disputed status rather than choosing an editorial winner without evidence. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 23: Cross-system export

Setup: A record moves from an international CMS into a legacy spreadsheet. The authoring job is to protect Unicode, separators, version and script metadata during export. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Check that import and re-export retain the original label. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

Transfer task 24: Security-sensitive lookalike

Setup: An account handle contains characters visually similar to another handle. The authoring job is to flag for identity validation rather than normalising away the difference. Start with the original name or identifier as received, not with a machine-generated approximation. Write a short memo identifying the referent, the output audience, the fields whose exact spelling is necessary and the fields that may be formatted for local presentation.

Acceptance test: Use a trusted channel before asserting the accounts match. Give a second reader the translated excerpt without the source and ask them to find the correct entity. Repeat with a compact headline or card. If the reader reaches a different record, if a private alias is disclosed, or if a model code or accession number changes, the translation fails regardless of grammar. A documented unresolved match is preferable to a confidently wrong one.

A release checklist for names and entities

Release gate 1: The actual person, work, place, organisation or object is identified by appropriate evidence, not inferred from spelling alone. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 2: The exact source label is retained somewhere in the authorised record, including script, punctuation and relevant diacritics. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 3: An approved target-language name or transliteration is distinguished from an automatic phonetic aid or speculative spelling. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 4: Where two entities share a spelling, the context supplies a reliable distinction or the identity remains unresolved. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 5: Where one entity has several names, the relationship is documented with appropriate provenance and privacy limits. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 6: Name parts are not inferred solely from their position or the number of spaces. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 7: Mononyms and compound or multiple surnames are preserved rather than forced into an unsuitable schema. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 8: Honorifics, offices, credentials and roles are not invented or merged into the personal name. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 9: Exact technical identifiers and model codes have been checked character by character. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 10: Historical names are not retroactively modernised inside quotations without editorial distinction. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 11: Search keys, sort keys and normalised matching forms are not confused with user-facing labels. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 12: Any script conversions use a documented method, with authorised exceptions where applicable. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 13: Mixed-script and confusable strings have received the appropriate security review for their use. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 14: Right-to-left display, punctuation and nearby numerical identifiers have been tested in the actual interface. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 15: Language tags are used for content or presentation metadata, never as a proxy for identity or nationality. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 16: All hyperlinks, headings, alt text, captions, citation tables and derivative summaries resolve to the intended referent. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 17: Privacy and consent decisions have been applied consistently to aliases and former names. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 18: A second reviewer has completed a target-only identification test for consequential records. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 19: Changes in approved names or mappings are versioned so a later edition can reproduce decisions. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Release gate 20: The approved publication is a translation of the intended entity relationships, not a new source of unverified identity claims. The evidence can be a verified original record, a source-author clarification, a known authority mapping or a documented editorial rule, depending on the risk. A reviewer should note who approved the decision and where it was applied. This converts a broad instruction to check names into a specific, repeatable test that another reviewer can reproduce.

Forty practical identity terms

Entity

A distinct person place organisation work product or other referent identified by a text. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Entity mention

A span of text referring to an entity in context. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Proper name

A conventional designation associated with a particular person place or thing. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Onomastics

The study of names and naming. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Exonym

A name used in another language for a geographic entity. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Endonym

A name used in a local language for a geographic entity. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Transliteration

A systematic representation of writing from one script in another. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Transcription

A representation often oriented toward pronunciation or actual speech. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Romanization

Representation of a name or text using Latin characters. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Phonetic guide

A reader aid for pronunciation that need not be official spelling. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Script

A system of writing represented by characters. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Language tag

A structured marker for language and sometimes script or regional usage. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Locale

A linguistic or cultural presentation context. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Alias

An alternative designation attached to an entity with relevant evidence. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Pseudonym

A chosen alternative name used in a public or creative identity. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Mononym

A name consisting of one personal-name element. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Given name

A personal-name component assigned to a person in some naming systems. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Surname

A family-name component in naming systems that use one. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Patronymic

A name element formed with reference to a parent. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Matronymic

A name element formed with reference to a mother. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Honorific

A form of respectful address or status marking. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Title

A word or phrase denoting rank function position or a work’s name depending on context. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Diacritic

A mark that modifies or distinguishes a written character. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Unicode

A standard for representing text characters across writing systems. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Normalization

A process defining a canonical or compatibility text representation. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

NFC

A Unicode canonical composition normalization form. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

NFKC

A Unicode compatibility composition normalization form. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Confusable

Characters or sequences with visually similar appearances. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Bidirectional text

Text mixing right-to-left and left-to-right writing directions. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Collation

Rules for sorting and comparing textual strings. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Sort key

A representation used to order records rather than necessarily to display a name. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Display label

The visible name chosen for a particular interface or publication context. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Identifier

A value intended to distinguish an object or record within a stated scheme. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Authority record

A record connecting an entity with documented identity data and names. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Entity resolution

The task of deciding which entity one or more mentions denote. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

False merge

An erroneous decision that two distinct entities are one. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

False split

An erroneous decision that one entity is represented as multiple entities. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Versioning

Recording which state of a record or mapping applies at a given time. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Provenance

Information about origin history and responsibility. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Privacy boundary

A rule limiting what identity-related data may be shared in a given context. In a translation project, identify whether this term describes the person or object, the visible writing of its name, or a processing rule used by software. The same word may appear in a source, a catalogue and a target-language note with different operational functions. Defining the function before translating prevents an indexing convenience from becoming an identity claim.

Frequently asked questions

1. Should proper names be translated or transliterated?

Choose the treatment from the actual entity, established target usage, source authority and reader purpose. A person’s name often remains a name; a place may have an established exonym; a book may have a published translated title. No blanket rule fits every case.

2. What is the difference between transliteration and translation?

Transliteration represents writing in another script under a mapping system. Translation conveys meaning between languages. A romanised name can still refer to the same person without translating the meaning of their name.

3. Can two romanised spellings identify the same person?

Yes, but the relationship should be supported by a verifiable record or direct author confirmation. Similarity alone is insufficient.

4. Can two people have the same written name?

Yes. Distinguish namesakes using source-specific evidence such as works, roles, dates or independently issued identifiers.

5. What if the person’s preferred spelling differs from an official document?

Keep context-specific fields separate. Use the verified form where the transaction requires it, and the preferred public form where appropriate and authorised.

6. Must every person have a first and last name?

No. Name structures vary widely, including single names and structures with several components. Forms and translated publications should not invent missing parts.

7. Should I remove accents to make names easier to search?

Do not remove accents from authoritative visible names without justification. A search system may maintain a separate matching key, but it should not overwrite the source.

8. Does Unicode normalization prove two names refer to the same person?

No. It can help compare text representations, not establish real-world identity. Human identity and string equivalence are different problems.

9. What is NFC?

NFC is a Unicode normalization form using canonical composition. The current Unicode normalization specification explains its properties; it is not a universal entity matching method.

10. Does a BCP 47 tag reveal a person’s nationality?

No. It describes language-related content and sometimes script or regional variation. It should not be used to infer citizenship or personal origins.

11. How should Arabic or Hebrew names be placed in English text?

Use appropriate directional isolation in supported markup and test actual rendering with surrounding numbers and punctuation. W3C internationalisation guidance explains common patterns.

12. What if a name is written in both Latin and another script?

Preserve both authorised forms and label their roles. Avoid treating a transliteration as a replacement for original-script evidence.

13. How do I translate a pen name?

Use the appropriate public author identity and approved title metadata. Do not expose a private civil identity merely because a separate source may contain it.

14. What if an organisation changed its name?

Verify whether the underlying entity continued or a different legal entity succeeded it. Date-scope the naming relation and preserve history.

15. Should I translate brand names?

Follow authorised brand usage and preserve exact model codes. Translate ordinary descriptive text separately.

16. Are scientific names the same as common names?

No. The scientific taxon can offer a more precise anchor, while common-name labels vary across languages and regions. Confirm the relevant taxonomic concept.

17. What if two books share a translated title?

Use author, publication, edition and reliable work identifiers to determine the reference. The target title alone is not conclusive.

18. Should historic place names be replaced by current names?

It depends on document purpose. Preserve historical quotations accurately and explain modern reference where appropriate rather than silently rewriting the original voice.

19. Can search engines resolve identity for me?

They can offer leads, but ranking and string similarity are not proof. Verify with context and appropriate source authority.

20. What is an authority record?

A structured reference that connects identity information and names. Its authority depends on scope and issuing institution; there is no single universal record for every entity.

21. How can a translator handle uncertain identification?

Record the plausible candidates, evidence and unresolved status. Ask the source owner or qualified reviewer rather than guessing.

22. What should a translation memory store about names?

Alongside approved renderings, keep entity references, context, label type, source evidence, restrictions and version information where supported.

23. Can a spelling correction change identity?

Yes. Especially with namesakes and model codes, a seemingly minor change can point to a different referent. Test the corrected target against the source record.

24. How do you protect private previous names?

Only use them when lawful and authorised for the purpose, with access controls and disclosure review. Public cross-references are not automatically justified.

25. Is the visually closest word always the best target name?

No. A familiar-looking approximation can be misleading. Prefer a well-evidenced identity relation and a context-appropriate display form.

26. Does the CLDR person-name standard identify people?

No. It concerns data structures and formatting of names, not independent proof of who a person is.

27. What about official identifiers?

Use them only within their valid issuing scheme and scope. Exact transcription and authorised access are important; never invent a supposedly global key.

28. How do I test a translated name?

Have another reader identify the intended entity using the target text, then compare the result with the original source evidence and permitted name variants.

29. Where should I learn about proper nouns?

The established Vocabulary proper-names guide explains naming and onomastics. This chapter focuses on preserving identity across translation and publication.

30. What is the final principle?

Change the label when justified by language and audience, but do not change the underlying referent or invent authority, history or verification.

Research references and terminology sources

Unicode CLDR 49, Part 8: Person Names (released 21 October 2026) describes cross-locale person-name structures, formatting, order, contexts and name parts. Crucially, formatting presumes supplied data; it is not a system for parsing arbitrary names or establishing legal identity.

Unicode CLDR 49, Part 1: Core (21 October 2026) covers locale data structure and identifiers. CLDR 49, Part 2: General covers display names and transforms. These sources guide localisation but do not authorise an unverified identity mapping.

Unicode Standard Annex #15: Unicode Normalization Forms (Unicode 18.0.0, 12 August 2026) explains NFC, NFD, NFKC and NFKD. Text normalisation can simplify string comparison, but it does not determine whether two names denote the same person or item.

Unicode Technical Standard #39: Unicode Security Mechanisms explains confusable text and identifier-security issues. Visual similarity across fonts and scripts cannot by itself validate the identity of an account, a domain or a name.

W3C Internationalization: Personal names around the world examines mononyms, patronymics, name order, different scripts, aliases and design risks in forms. The examples illustrate broad variation; a person-specific field still requires appropriate source evidence.

W3C: Inline markup and bidirectional text in HTML explains text-direction isolation and how mixed-script phrases can reorder punctuation and adjacent strings. The advice supports robust presentation; it does not validate the identity of the phrase’s subject.

IETF RFC 5646: Tags for Identifying Languages (September 2009, part of BCP 47) specifies language tags and optional subtags including script and region. Such tags identify linguistic content and are not person or organisation identifiers.

Master Art of Translation | The Entity Identity Preservation System — How Names, Scripts, Aliases and Stable Identifiers Stay Connected