Translating a BIM data template or construction asset property dictionary demands more than finding the nearest word in another language. A property label may change for a French or Japanese asset manager while its underlying stable identifier, unit, permissible value and link to an asset class must remain consistent. If a translated field quietly changes what measurement it represents, software can ingest a syntactically valid file and still deliver the wrong operational meaning. This is a focused guide to translating machine-readable BIM asset data, not a general introduction to translation.
ISO 23387:2025 provides the current framework for data templates for objects used throughout the asset life cycle. The official ISO catalogue describes machine interpretability, data dictionary maintenance, links with classification systems and an XML Schema Definition; it also notes that the standard does not provide the content of any particular template. ISO 23387:2025 superseded the 2020 edition in September 2025. We use that verified scope to explore practical multilingual property naming, units, version control and handover QA without reproducing paid clauses.
Quick answer: localise user-facing property labels and explanatory definitions where appropriate, but preserve approved property identifiers, classification references, data types, unit constraints and enumerated codes. Record the exact data-template version and governing dictionary. Distinguish a missing value from zero or not applicable. A bilingual reviewer should be able to reconstruct the same asset fact in both languages and show that the translated field still binds to the correct machine-readable property.
In three minutes: the practical answer
- Identify the document, the competent owner and the exact published edition before translating.
- Lock all identifiers, codes, sample IDs, units, dates and controlled status fields.
- Separate observation, interpretation, requirement, prohibition and authorisation.
- Draft readable English explanations of difficult concepts before choosing a target-language equivalent.
- Perform linguistic review and a second, domain-specific technical review.
- Record open questions, approvals and release versions; do not conceal unresolved evidence.
What the official standard covers—and what this guide does not
ISO 23387:2025 concerns machine-interpretable data templates for objects used throughout an asset’s life cycle, linked to data dictionaries and classification systems. The second edition published in September 2025 replaced the withdrawn ISO 23387:2020 edition. It includes an XML Schema Definition and does not supply the actual content of any data templates. This scope summary is grounded in the publicly accessible ISO catalogue record. The full standard is proprietary and may include more detailed technical conditions, amendments and references than its public abstract. Do not treat the following fictional cases as formal ISO requirements or as proof that a specific product, project or laboratory meets the standard. They illustrate language decisions and evidence handling only.
The exact search question addressed here is: How do multilingual BIM teams translate asset data template property names and descriptions without changing identifiers, enumerated values, units, classification links or machine-readable exchange meaning? This is a specialist educational article for BIM coordinators, digital construction teams, asset owners, data dictionary maintainers and technical localisation specialists. Its task is narrower than teaching translation from the beginning, replacing the official standard or providing regulated technical instructions. At each stage, the translator should establish whether the source is a definition, a requirement, a test finding, a sales claim or a historical record. That classification controls what the target-language reader is entitled to believe.
A realistic translation job: establish the evidence pack
Imagine receiving a bundle containing a revised procedure, an earlier supplier quotation, a spreadsheet with partly translated column names, a PDF diagram, a machine-generated translation memory and an email requesting a same-day publication. The order of work matters. First identify the authoritative document and its revision, the target markets and actual audience. Then separate controlled identifiers from localisable display labels. Extract units, negations, modal verbs, qualifiers and references into a review sheet. Finally, compare the release candidate with the source in the format in which a reader will actually use it.
A translation memory can suggest words but does not decide which source is current. A bilingual professional can judge grammar but cannot assume responsibility for an undocumented engineering or laboratory claim. A technical reviewer can approve domain meaning but may not notice a target-language ambiguity that a native reader catches immediately. For that reason the most robust workflow has independent checks, explicitly described authority and a retained change log. Every significant question should be answerable by a cited source field, an approved clarification or an appropriately cautious statement that no evidence has yet been supplied.
The important concepts, one by one
1. A label is not a property identifier
A data dictionary can use a stable machine identifier for a property while presenting different human-readable labels in different languages. The identifier, not the translated wording, maintains a link to the defined concept. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A software importer may mis-map an asset if the display name is treated as the unique key. Consider a constructed situation: A property appears as Net area to readers but has an independent dictionary identifier. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to keep internal identifiers fixed and record approved multilingual labels as language-specific annotations.. The document owner should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
2. A property definition sets the concept boundary
Two terms that look synonymous in everyday English may denote different measurements, physical states or lifecycle stages in a building asset. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: Changing an operational design property to a measured-as-built value changes the information need. Consider a constructed situation: A database separates nominal design airflow from measured commissioning airflow. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to use definitions, units and evidence source to validate the concept before selecting target wording.. A receiving technical team should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
3. The object class gives a name its meaning
The same noun can describe a door, a pump, a space, a network or a complete building. Its correct translation depends on what type of object the template represents. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A translated property might be wrongly attached to a component rather than an assembly. Consider a constructed situation: A building operator sees a property linked to an air-handling unit and a room. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to keep object class and inheritance links visible in the translation work package.. A bilingual reviewer should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
4. Data types are not cosmetic
Numbers, strings, booleans, dates, enumerations and references behave differently in a machine-readable schema. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: Changing a boolean false into a free-text No without respecting the schema can break validation. Consider a constructed situation: One export holds a numeric efficiency value and another stores a descriptive performance category. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to localise presentation but keep the underlying data type and valid machine value unchanged.. The person implementing the translated instruction should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
5. Units, decimal signs and scale must be governed
Asset data may use dimensions such as metres, millimetres, square metres, litres per second or kilowatts. The presentation locale may change punctuation, but an authorised conversion needs explicit rules. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A decimal separator or mistaken unit scale can create a thousandfold difference. Consider a constructed situation: A fan has a flow rate expressed in L/s and the receiving system expects m³/s. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to retain the source unit and conversion provenance; independently verify numeric transformations.. The document owner should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
6. Enumerations are finite decision sets
A pick-list may store a code and show a localised label such as open, closed, unknown or not applicable. The labels must never be allowed to redefine the accepted values. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A translator may merge unknown and not applicable because both sound like absence. Consider a constructed situation: A door state is an enum, whereas a maintenance comment is unrestricted text. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to keep each code paired to one approved concept with clear usage examples.. A receiving technical team should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
7. Versioned dictionaries must remain referentially stable
Asset information can live for decades while classification systems, software and data dictionaries evolve. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A translated property may refer to a newer definition than its source model. Consider a constructed situation: A maintenance platform imports a template updated after original construction handover. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to record dictionary version, effective date and migration mapping instead of silently substituting a recent term.. A bilingual reviewer should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
8. Relationships are part of the meaning
An asset may be part of an assembly, located in a space, serving another system or associated with a document. A name alone may not show the link. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: Translating serves as located in changes the relationship graph. Consider a constructed situation: A ventilation component has both location and functional-service properties. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to review relation types and directions as carefully as labels.. The person implementing the translated instruction should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
9. Lifecycle status separates planned and measured data
Concept, design, procurement, construction, commissioning, operation and decommissioning can produce different values for a similar property. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: As designed should not become verified after installation. Consider a constructed situation: An asset register includes design capacity and latest inspected capacity. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to translate lifecycle qualifiers and the date, owner and evidence for each value.. The document owner should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
10. Classification links are not free keywords
A classification reference is a governed mapping to a system, not a casual topical tag. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A translated class may sound more familiar while pointing to the wrong category. Consider a constructed situation: An IFC class and a national classification code appear beside a data template. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to preserve source codes and apply approved crosswalks rather than rewriting them.. A receiving technical team should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
11. XML and schema constraints need dual review
Machine-readable template exchange may depend on character encoding, required fields, namespaces, schemas and valid value types, independent of human prose. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A label with an unescaped character might disrupt parsing while its language is correct. Consider a constructed situation: A software exporter localises a textual description in a structured payload. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to run schema validation plus bilingual concept review; neither replaces the other.. A bilingual reviewer should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
12. Missing values are not interchangeable
A blank field can mean unknown, not measured, withheld, inapplicable or a true empty string, depending on the schema. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: Turning missing into zero may falsely report an asset with no capacity. Consider a constructed situation: A system has not received a commissioning value for an installed device. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to preserve null semantics, required/optional status and any explanatory missing-value reason.. The person implementing the translated instruction should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
13. Templates do not guarantee valid observations
A well-structured data template can still carry a wrong measurement, an obsolete product claim or an unapproved assumption. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: A schema-valid translation could be mistaken for validated engineering evidence. Consider a constructed situation: A field passes XML validation but its evidence document describes a different model. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to check both technical source fidelity and provenance of the supplied value.. The document owner should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
14. Ownership and governance are long-term requirements
Different teams may own the template schema, glossary, asset records and translated explanations. Without clear responsibilities, small local edits become permanent divergences. In a real translation request, the source may be an annotated drawing, a database field, a controlled procedure or a short excerpt copied into an email. Those formats are not interchangeable: a title often omits the condition supplied in a table, whereas a narrative report may refer to a document version already superseded. Before drafting target-language words, locate the evidence that defines the term and the document’s intended use. This preserves the distinction between explaining what a record says and independently deciding that the record is correct.
The practical trap is this: An ungoverned synonym in one language might gradually become a separate data concept. Consider a constructed situation: A facilities team requests a clearer label after operators misunderstand a field. It is tempting to repair an awkward source by picking the most common dictionary equivalent, but that could change the subject, scope, time, number or obligation. A fluent sentence can therefore be less accurate than a cautious one that preserves the technical restriction. Document the uncertainty at the point where it occurs; asking a targeted question is better than scattering vague caution throughout the final text. For world-facing readers, avoid assuming the practices of one country automatically govern another.
A useful review action is to route changes through dictionary governance, log review decisions and release mapped versions.. A receiving technical team should compare the source statement with the proposed target language, then point to the same named object and the same evidence after reading both. Ask specifically what the reader would do next, whether that action is authorised, and which words tell them a result is observed rather than inferred. Where the document makes no claim about a circumstance, the translation should not fill the silence. Record the source version, termbase decision and reviewer comment so a later revision can be checked without repeating the whole argument.
Twenty-plus original worked translation decisions
These cases are simplified and hypothetical. Each one presents an English source fragment or record label and a target-language meaning brief rather than pretending that a single literal translation works in all languages. The goal is to preserve the information that a qualified translator must express. The eventual French, Spanish, Arabic, Mandarin or other phrasing should be chosen by a competent linguist and reviewed in its target context. None of the sample labels certifies a real result.
Case 1: Nominal airflow
Constructed source situation: Design property on a ventilation component. The source wording is “Nominal airflow”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Confusing planned capacity with measured performance. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Preserve nominal and link the unit to the correct property identifier. For a bilingual quality reviewer, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 2: Measured airflow
Constructed source situation: Commissioning evidence records an observed value. The source wording is “Measured airflow”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Renaming it nominal because the number resembles a design value. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep measurement status, date and measurement provenance. For the receiving practitioner, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 3: Net floor area
Constructed source situation: Area of a defined asset or space. The source wording is “Net floor area”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Replacing net with gross. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Check the authoritative measurement definition and space boundary. For a document controller, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 4: Door fire-resistance classification
Constructed source situation: Product property linked to a controlled class. The source wording is “Door fire-resistance classification”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Turning a classification into an unsupported time guarantee. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep the class code and applicable source record. For a non-specialist reader, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 5: Asset identifier
Constructed source situation: Stable record key across systems. The source wording is “Asset identifier”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Translating or reformatting the machine key. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Localise the label; preserve identifier bytes. For a bilingual quality reviewer, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 6: Property description
Constructed source situation: Human-language explanation of a governed field. The source wording is “Property description”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Paraphrasing beyond the dictionary definition. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep concept boundaries and leave codes untouched. For the receiving practitioner, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 7: Boolean false
Constructed source situation: Machine representation of a yes/no field. The source wording is “Boolean false”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Using a translated string where boolean is required. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Map user-facing text to the unchanged boolean. For a document controller, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 8: Unknown state
Constructed source situation: An enumeration expresses unavailable status. The source wording is “Unknown state”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Treating unknown as false. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Preserve the unknown enum and reason for missing evidence. For a non-specialist reader, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 9: Not applicable
Constructed source situation: An enum indicates a property is irrelevant. The source wording is “Not applicable”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Rendering as not yet tested. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Distinguish inapplicability from unverified applicability. For a bilingual quality reviewer, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 10: 3.5 kW
Constructed source situation: Power value with specified unit. The source wording is “3.5 kW”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Moving decimal or scaling incorrectly. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Preserve numerical meaning and authorised unit conversion. For the receiving practitioner, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 11: 1200 mm
Constructed source situation: Dimensional property in product records. The source wording is “1200 mm”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Mistaking length for area. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep quantity kind, unit and original precision. For a document controller, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 12: Part of assembly
Constructed source situation: Relationship from component to parent. The source wording is “Part of assembly”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Reversing parent-child direction. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Verify graph roles and target identifiers. For a non-specialist reader, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 13: Serves zone 4
Constructed source situation: Functional relationship from equipment to space. The source wording is “Serves zone 4”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Replacing serves with located in. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Preserve exact relationship and zone code. For a bilingual quality reviewer, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 14: As-designed value
Constructed source situation: A design-stage capacity field. The source wording is “As-designed value”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Implying physical verification. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep planned status and associated evidence. For the receiving practitioner, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 15: As-built value
Constructed source situation: A post-construction value with provenance. The source wording is “As-built value”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Assuming equivalence with commissioned test data. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Preserve evidence stage and revision. For a document controller, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 16: IFC class reference
Constructed source situation: A system uses a classification link. The source wording is “IFC class reference”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Localising or altering an official code. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep classification code and scheme name. For a non-specialist reader, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 17: Dictionary version 2.1
Constructed source situation: A data model is versioned. The source wording is “Dictionary version 2.1”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Dropping the suffix when exporting. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep version and migration requirements. For a bilingual quality reviewer, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 18: Deprecated field
Constructed source situation: An older property is still in some records. The source wording is “Deprecated field”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Deleting it from language files without mapping. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Preserve deprecation status and crosswalk. For the receiving practitioner, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 19: Unescaped ampersand
Constructed source situation: An XML-bound label contains a character requiring encoding. The source wording is “Unescaped ampersand”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Passing plain text straight into a serializer. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Use approved escaping and validate schema. For a document controller, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 20: Required attribute missing
Constructed source situation: A validation issue identifies an absent field. The source wording is “Required attribute missing”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Filling the gap with invented values. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep absence explicit and request authorised data. For a non-specialist reader, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 21: Display name
Constructed source situation: A UI may show a localised title. The source wording is “Display name”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Using translated title as the database key. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Separate presentation from identity. For a bilingual quality reviewer, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Case 22: Operating manual reference
Constructed source situation: A property links to a controlled document. The source wording is “Operating manual reference”. This phrase is a teaching example, not a quotation from the paid ISO standard. Its immediate surrounding text matters because a terminology database alone cannot tell whether the phrase is a property label, an instruction, an observation, an exception or a decision already signed off.
Risk of mistranslation: Replacing reference with a summary that loses traceability. That shift would make the receiving version answer a different question from the original. For example, the recipient might associate an identifier with the wrong item, interpret a condition as a universal permission, or treat a provisional statement as a completed technical conclusion. The relevant audit is semantic: compare who is acting, which item is affected, what state is claimed, whether any conditions limit it and what evidence supports the claim.
Meaning-preserving translation brief: Keep document ID and version with the translated description. For the receiving practitioner, ask for the authorising record rather than guessing from similar phrases in unrelated files. Retain all explicit negation, units, model numbers, classifications, date roles, qualifications and release states relevant to this example. A revised target sentence should be back-explainable without adding a new promise: another reviewer ought to reconstruct the same practical fact and know what remains unproved. If the underlying source is internally contradictory, mark the query unresolved and do not publish an invented harmonisation as a fact.
Four integrated practice dossiers
Dossier 1: a document crosses an organisational boundary
A fictional international project sends a translated pack covering “Nominal airflow”, “Measured airflow”, “Net floor area”, “Door fire-resistance classification”, “Asset identifier”. The receiving organisation has no direct access to the author’s internal discussions and sees only the released file. First mark where one line identifies an object, where another line is a result, and where a third line qualifies a condition or records an approval. The chief challenge is not translating each label in isolation; it is making sure their relationship remains intact. A version mismatch could make the combined pack appear internally consistent when its components actually describe different moments in the work.
The reviewer should apply five focused tests. Does the translation retain the same named thing? Does it distinguish measured information from proposed action? Does it preserve what is not known or still pending? Are numerical values, identifiers and coded states carried through without unauthorised changes? Can a receiving team trace the final sentence to a record rather than an assumption? Relevant resolution rules include: Preserve nominal and link the unit to the correct property identifier; Keep measurement status, date and measurement provenance; Check the authoritative measurement definition and space boundary; Keep the class code and applicable source record; Localise the label; preserve identifier bytes. These decisions need not make the target prose cumbersome; a brief note beside an ambiguous label may be more helpful than an unqualified rewrite.
Now imagine the recipient asks whether the entire dossier can be treated as final. A responsible translator does not answer by inspecting how confident the language sounds. The answer must come from issue status, revision references, approval signatures and the actual provenance of included evidence. If the source is a draft, translate it as a draft. If a test is pending, the target must communicate that it is pending. If a code has no approved definition, raise a query. This exercise trains a practical skill: preserve the precise boundaries that permit a future reader to make a sound decision without asking the translator to become the approving authority.
Dossier 2: a document crosses an organisational boundary
A fictional international project sends a translated pack covering “Property description”, “Boolean false”, “Unknown state”, “Not applicable”, “3.5 kW”. The receiving organisation has no direct access to the author’s internal discussions and sees only the released file. First mark where one line identifies an object, where another line is a result, and where a third line qualifies a condition or records an approval. The chief challenge is not translating each label in isolation; it is making sure their relationship remains intact. A version mismatch could make the combined pack appear internally consistent when its components actually describe different moments in the work.
The reviewer should apply five focused tests. Does the translation retain the same named thing? Does it distinguish measured information from proposed action? Does it preserve what is not known or still pending? Are numerical values, identifiers and coded states carried through without unauthorised changes? Can a receiving team trace the final sentence to a record rather than an assumption? Relevant resolution rules include: Keep concept boundaries and leave codes untouched; Map user-facing text to the unchanged boolean; Preserve the unknown enum and reason for missing evidence; Distinguish inapplicability from unverified applicability; Preserve numerical meaning and authorised unit conversion. These decisions need not make the target prose cumbersome; a brief note beside an ambiguous label may be more helpful than an unqualified rewrite.
Now imagine the recipient asks whether the entire dossier can be treated as final. A responsible translator does not answer by inspecting how confident the language sounds. The answer must come from issue status, revision references, approval signatures and the actual provenance of included evidence. If the source is a draft, translate it as a draft. If a test is pending, the target must communicate that it is pending. If a code has no approved definition, raise a query. This exercise trains a practical skill: preserve the precise boundaries that permit a future reader to make a sound decision without asking the translator to become the approving authority.
Dossier 3: a document crosses an organisational boundary
A fictional international project sends a translated pack covering “1200 mm”, “Part of assembly”, “Serves zone 4”, “As-designed value”, “As-built value”. The receiving organisation has no direct access to the author’s internal discussions and sees only the released file. First mark where one line identifies an object, where another line is a result, and where a third line qualifies a condition or records an approval. The chief challenge is not translating each label in isolation; it is making sure their relationship remains intact. A version mismatch could make the combined pack appear internally consistent when its components actually describe different moments in the work.
The reviewer should apply five focused tests. Does the translation retain the same named thing? Does it distinguish measured information from proposed action? Does it preserve what is not known or still pending? Are numerical values, identifiers and coded states carried through without unauthorised changes? Can a receiving team trace the final sentence to a record rather than an assumption? Relevant resolution rules include: Keep quantity kind, unit and original precision; Verify graph roles and target identifiers; Preserve exact relationship and zone code; Keep planned status and associated evidence; Preserve evidence stage and revision. These decisions need not make the target prose cumbersome; a brief note beside an ambiguous label may be more helpful than an unqualified rewrite.
Now imagine the recipient asks whether the entire dossier can be treated as final. A responsible translator does not answer by inspecting how confident the language sounds. The answer must come from issue status, revision references, approval signatures and the actual provenance of included evidence. If the source is a draft, translate it as a draft. If a test is pending, the target must communicate that it is pending. If a code has no approved definition, raise a query. This exercise trains a practical skill: preserve the precise boundaries that permit a future reader to make a sound decision without asking the translator to become the approving authority.
Dossier 4: a document crosses an organisational boundary
A fictional international project sends a translated pack covering “IFC class reference”, “Dictionary version 2.1”, “Deprecated field”, “Unescaped ampersand”, “Required attribute missing”. The receiving organisation has no direct access to the author’s internal discussions and sees only the released file. First mark where one line identifies an object, where another line is a result, and where a third line qualifies a condition or records an approval. The chief challenge is not translating each label in isolation; it is making sure their relationship remains intact. A version mismatch could make the combined pack appear internally consistent when its components actually describe different moments in the work.
The reviewer should apply five focused tests. Does the translation retain the same named thing? Does it distinguish measured information from proposed action? Does it preserve what is not known or still pending? Are numerical values, identifiers and coded states carried through without unauthorised changes? Can a receiving team trace the final sentence to a record rather than an assumption? Relevant resolution rules include: Keep classification code and scheme name; Keep version and migration requirements; Preserve deprecation status and crosswalk; Use approved escaping and validate schema; Keep absence explicit and request authorised data. These decisions need not make the target prose cumbersome; a brief note beside an ambiguous label may be more helpful than an unqualified rewrite.
Now imagine the recipient asks whether the entire dossier can be treated as final. A responsible translator does not answer by inspecting how confident the language sounds. The answer must come from issue status, revision references, approval signatures and the actual provenance of included evidence. If the source is a draft, translate it as a draft. If a test is pending, the target must communicate that it is pending. If a code has no approved definition, raise a query. This exercise trains a practical skill: preserve the precise boundaries that permit a future reader to make a sound decision without asking the translator to become the approving authority.
Practical translation QA before publication
- Source freeze: identify the file names, effective versions, approval source and revision history.
- Field locking: lock identifiers, date roles, units, codes and table columns against accidental transformation.
- Terminology review: explain each important concept in plain language, then apply the approved equivalents in context.
- Clause and condition review: check negation, modality, scope, exceptions, status and whether a statement is observed or authorised.
- Layout review: inspect line breaks, labels, tables, bidirectional scripts, special characters and truncated text.
- Technical sign-off: ask competent owners to verify claims, constraints and factual context.
- Release archive: keep source-to-target version mapping, resolved questions and the final approved files.
A good final test is to hand only the translated document to someone who did not participate in drafting. Ask them to identify the named item, the exact status or measurement, the evidence supporting it and any restriction on acting. If their answer differs from what a source-language reader would reconstruct, the translation is not ready even if it has no grammatical errors. The check must be done with realistic document placement, because column widths, footnotes, symbol placement and small-print warnings can change practical comprehension. Keep any review findings attached to the release record.
Questions readers ask
Is ISO 23387:2020 still current?
No. ISO records the 2020 edition as withdrawn and replaced by ISO 23387:2025, published in September 2025. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
Can translators change an internal property ID?
Not merely for readability. Stable identifiers are governed by the dictionary and systems integrating with it. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
Is an XML-valid file necessarily semantically correct?
No. Valid syntax does not establish that a property’s concept, measurement or classification mapping is accurate. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
Should values be converted to local units?
Only under approved project rules with a verified numeric transformation and retained source provenance. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
Can unknown and not applicable be combined?
No, unless a governing schema explicitly defines them as the same concept; they normally answer different questions. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
Do data templates themselves provide engineering measurements?
No. The ISO catalogue says the standard sets the data-template concept and structure, not the contents of individual templates. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
Who should approve a label change?
The dictionary owner or delegated domain expert, supported by bilingual and technical review. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
What is the best release test?
A translated record must still resolve to the same identifier, concept, type, unit, classification and source evidence. The responsible review should compare the current controlled source and the actual target-market use rather than assume an isolated sentence provides every fact. If a material point remains unresolved, communicate its status to the document owner before release.
A four-week learning and quality-improvement route
Week one: select a short authorised sample and build a term register with the source wording, concept definition, context, prohibited false friends and target-language decision. Ask a colleague to explain the intended reader action in ordinary words. Week two: build paired source and target table views and compare the identifiers, units, dates and document versions. Use intentionally ambiguous fictional test rows to train the habit of raising explicit questions rather than filling gaps silently.
Week three: conduct a full bilingual review with a separate subject-matter reviewer. Give each reviewer distinct tasks, because the same person is less likely to notice assumptions introduced during drafting. Check rendering in PDF, web and printed layouts as relevant. Week four: simulate a release change: revise a source qualifier, update only affected target passages, mark the old version superseded and audit the archive. Successful learning means the translated documents remain usable after an update, not simply that an initial paragraph reads fluently.
Official source and the established eduKate learning pathways
Primary reference: ISO 23387:2025 — official ISO catalogue. This independent article is explanatory commentary using original constructed examples. It does not reproduce the purchased standard, offer a certificate of conformity, prescribe a clinical or industrial operation, or override the responsible professional and applicable jurisdiction. Readers should consult the authorised full edition, their own contracts and qualified advisers before applying technical decisions.
For the broad architecture of moving meaning between languages, see Master Art of Translation — The Complete System for Moving Meaning Between Languages (EDKSG-TRANS-MASTER-WORLD-000). For building accurate conceptual knowledge, use the protected Vocabulary Learning Hub. For English syntax, modality, reference and meaning across languages, use How English Works — Translation and Meaning Across Languages. These resources retain ownership of foundational learning; this article solves only the specific bim asset data templates translation problem.
The practical standard for a successful translation is simple to describe and demanding to meet: a responsible person in another language should recognise the same object, the same evidence, the same uncertainty and the same authorised next step. When those elements travel intact, the text is useful. When any of them changes silently, elegant language can become a source of error.