Master Art of Translation | The TBX 3.0 and ISO 30042 Termbase Exchange System — How Concepts, Dialects, Preferred Terms, Sources and XML Validation Keep Their Meaning

How do you translate a TBX 3.0 termbase, move terminology between computer-assisted translation systems, and validate an ISO 30042:2019 terminology exchange without losing concepts, preferred terms, forbidden terms, language tags or the evidence supporting each decision? The answer is to treat TermBase eXchange (TBX) as a structured exchange contract rather than a bilingual spreadsheet: preserve the identity and boundaries of each concept, declare the actual TBX dialect, keep descriptive and administrative information at the right level, and verify what the receiving system actually imported. This guide explains that system for translators, terminologists, developers, localization teams and reviewers.

The hidden problem is not whether a translated word looks plausible. A file can be well formed XML, contain apparently reasonable translations and still be unusable because two concepts were merged, a deprecated designation was promoted, a source citation vanished, a language variant was mislabelled, or an importer silently discarded a data category it did not understand. A file that opens successfully has passed only one test. It has not proved semantic equivalence, safe usage, correct term status, reliable provenance, or round-trip preservation.

Quick route: identify the exact source and target dialects; inventory concept identifiers and language sections; classify each field as translatable, controlled, immutable or review-only; keep preferred, admitted and deprecated status distinct; validate XML structure and permitted data categories; inspect a sample in the receiving termbase; compare exported and re-imported concepts; investigate losses before release. This article is the specialist interchange and validation child of the Master Art of Translation root. It does not replace the wider terminology system or the vocabulary and English learning owners.

• Search task: how to translate, export, import, migrate and validate a TBX 3.0 terminology database safely.
• Primary distinction: a concept is not the same thing as its spelling, language label, synonym or exported row.
• Technical boundary: ISO 30042:2019 defines the TBX framework and TBX-Core; a chosen dialect determines additional allowable data categories and constraints.
• Core deliverable: a documented import/export contract with a preservation report, not merely a file that parses.
• Release rule: structural validation plus semantic review plus receiving-tool inspection plus reversible evidence.

1. Search intent: what people really mean by ‘translate a TBX file’

A reader asking how to translate a TBX file may mean at least four different jobs. One is adding legitimate target-language terms to concepts already established in a termbase. Another is migrating an existing multilingual termbase between products. A third is mapping a proprietary glossary into a standard interchange format. A fourth is troubleshooting a schema or import failure. These tasks share terminology, but their acceptance criteria differ. Adding a translation requires linguistic and domain review; migrating records requires field fidelity; creating XML requires dialect awareness; repairing imports requires comparing the intended semantic record with the application state. Begin by naming the actual job.

TBX is not a translation engine. It is a family of structured representations for terminology data. A translation-memory file normally stores reusable text segments, whereas a terminological concept entry answers a different question: what concept is meant, how is it designated in different languages, and what evidence or conditions govern those designations? Consequently, automatically aligning two columns of words does not produce a reliable multilingual terminology collection. It may produce useful candidates, but the candidates need their own senses, domain boundaries and approvals.

ISO 30042:2019 is the published second edition of the TBX standard. Its public ISO abstract explains the metamodel, data categories, the data-category-as-attribute and data-category-as-tag XML styles, and dialect-definition methodology. TBX Info describes version 3 as published in 2019 and reports that a standards review is under way; that review is not a reason to describe a draft revision as an already-published replacement. ISO 16642:2025 supplies a newer terminology-markup framework, and ISO 704:2022 sets terminology-work principles. These documents have related but different roles.

2. The mechanism: separate meaning, language, term and exchange format

Consider the word ‘cell’. In a research report it might denote a biological unit, a spreadsheet intersection, an electrical component or a prison room. An export that treats ‘cell’ as a unique concept key cannot safely move the concept through another language. The term is surface language; the concept is the unit of knowledge. A concept-oriented termbase can hold distinct entries for the different senses, each with definitions, subject fields and terms in appropriate languages. The exchange format must preserve that separation rather than flattening it into a convenient word list.

Similarly, ‘preferred’ does not mean the same as ‘most frequent’, and ‘deprecated’ does not always mean incomprehensible. Preferred is a controlled recommendation within a policy and domain; a deprecated form may remain searchable so historic documents can be interpreted; a forbidden form may be blocked in new authoritative materials. These are editorial and governance decisions. A transfer that strips status changes the behaviour of writers and translators even if the wording itself survives.

A robust TBX workflow therefore combines three independent checks. Structural correctness asks whether the instance is valid under the selected dialect, XML namespaces and constraints. Semantic correctness asks whether concepts, definitions, language assignments and status decisions still mean what they meant. Operational correctness asks whether the receiving application can display, search, enforce and re-export the intended information. A pass in one layer never automatically proves a pass in the others.

3. Repeatable end-to-end method

• 1. Define the request: Write down whether the job is term creation, language expansion, dialect conversion, platform migration or repair.
• 2. Freeze input: Export the source unmodified and record its origin, version, hash, date and permissions.
• 3. Discover dialect: Identify the source TBX version and declared dialect rather than guessing from the .tbx extension.
• 4. Inventory concepts: Count concept entries, language sections, term records, status values, links and source notes.
• 5. Define field map: Mark each field as content, code, identifier, metadata, controlled value or unrecognized extension.
• 6. Establish concepts: Distinguish homonyms and related but non-equivalent concepts before translating terms.
• 7. Assign languages: Use correct language tags; do not silently label a regional term with a generic tag.
• 8. Translate evidence: Translate definitions, contexts and labels only when the receiving audience needs them and the field permits it.
• 9. Approve term status: Set preferred, admitted, deprecated or other supported statuses using the relevant project policy.
• 10. Preserve references: Keep bibliographic and internal links traceable; investigate unresolved references.
• 11. Validate structure: Check well formed XML, namespaces, dialect schema and relevant semantic constraints.
• 12. Inspect import: Open representative records in the actual target termbase, not just in an XML editor.
• 13. Round trip: Export again and compare concepts and fields, allowing documented normalization only.
• 14. Investigate losses: Classify every dropped or transformed field and its effect on authoring decisions.
• 15. Release with evidence: Publish a signed-off field map, warnings log, reviewer decisions and rollback package.
• 16. Maintain governance: Schedule changes in terminology, product names, subject taxonomies and import-tool behaviour.

The gates are a practical production method, not a claim that ISO prescribes a sixteen-step workflow. Teams can adapt the control count, but they should not discard the underlying proof obligations. The decisive question is always what a later reader, translator, CAT-tool suggestion engine or software product will do because the record has changed. If a field influences usage or trust, its transformation deserves an explicit review.

4. Structural concepts and controls: forty-six detailed mechanisms

1. Concept entry identity

What this mechanism controls. Concept entry identity concerns one concept distinct from every term used for it. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a label is mistakenly used as a globally unique identifier. Consider the specific example of two identical terms describing different concepts. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve a stable concept ID together with definition and subject field. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. compare identity count and definition scope before and after import. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

2. Concept scope note

What this mechanism controls. Concept scope note concerns the boundary between a concept and near neighbours. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a short definition fails to exclude a neighbouring sense. Consider the specific example of maintenance stop confused with planned outage. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve a scope note describing inclusion and exclusion tests. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. ask an expert to apply both labels to a boundary case. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

3. Definition placement

What this mechanism controls. Definition placement concerns what is defined at concept level rather than term level. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a definition is copied under a single language and disappears elsewhere. Consider the specific example of a multilingual record containing inconsistent explanations. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve an approved concept definition and its language provenance. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. verify each target-language definition expresses the same concept. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

4. Language section

What this mechanism controls. Language section concerns the association of one language with its term designations. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, terms migrate under an incorrect language tag. Consider the specific example of a regional variant imported under generic language. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve a declared language identifier consistent with the actual text. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test sorting, search and language filtering in the receiver. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

5. Term section

What this mechanism controls. Term section concerns a designation and its term-specific annotations. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, one status accidentally applies to all synonyms. Consider the specific example of acronym marked preferred while its full form is admitted. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve each term entry with its distinct type, status and evidence. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. inspect the status of every individual term. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

6. Preferred designation

What this mechanism controls. Preferred designation concerns the authorized wording for new controlled writing. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, frequency is mistaken for policy preference. Consider the specific example of a common legacy phrase incorrectly promoted. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve an explicit editorial decision with effective date. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. check the authoring suggestion returned for a new sentence. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

7. Admitted designation

What this mechanism controls. Admitted designation concerns an acceptable alternative without default preference. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, the importer promotes permitted synonyms as equal defaults. Consider the specific example of two accepted short forms used in separate contexts. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the distinction between approval and default recommendation. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. verify the term-picker order and any policy labels. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

8. Deprecated designation

What this mechanism controls. Deprecated designation concerns a legacy term retained for interpretation but not recommended. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, deprecation is stripped on import. Consider the specific example of an obsolete equipment nickname returned as fresh copy. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the searchability of history alongside a caution against new use. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. request a new-document suggestion and a historical search. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

9. Forbidden designation

What this mechanism controls. Forbidden designation concerns a term the governance team prohibits in a defined context. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a forbidden status is mapped to ordinary alternate label. Consider the specific example of a misleading clinical abbreviation moved into a public app. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the restriction reason, scope and editorial authority. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test negative terminology QA and human exception handling. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

10. Subject field

What this mechanism controls. Subject field concerns the domain in which a term has a defined meaning. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, two fields with overlapping labels are merged. Consider the specific example of ‘port’ in networking and marine logistics. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the domain taxonomy value and concept association. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. search separately by discipline and examine disambiguation. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

11. Context quotation

What this mechanism controls. Context quotation concerns a sentence demonstrating how a term works in use. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, the quotation is mistaken for a definition or normalized into false grammar. Consider the specific example of a sentence showing ‘charge’ as fee rather than electrical quantity. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve authentic surrounding syntax plus source attribution. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test whether usage guidance survives and remains legible. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

12. Part of speech

What this mechanism controls. Part of speech concerns a grammatical classification attached to the proper designation. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, noun status is inherited by a verb homograph. Consider the specific example of ‘record’ noun and ‘record’ verb. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the part-of-speech category as used by the dialect. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. ask the tool to filter by grammatical role. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

13. Grammatical gender

What this mechanism controls. Grammatical gender concerns relevant nominal agreement information for a language. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, gender is generalized to all languages or stripped as decoration. Consider the specific example of the feminine noun variant used with masculine agreement. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve a language-specific annotation attached to the right term. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test grammatical selection with a complete example sentence. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

14. Geographical usage

What this mechanism controls. Geographical usage concerns the region or variety in which wording is used. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a territory tag is treated as a universal concept property. Consider the specific example of regional automotive vocabulary with same underlying concept. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the permitted-use scope of a designation. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test audience-specific selection without splitting concept identity. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

15. Term type

What this mechanism controls. Term type concerns whether a designation is full form, abbreviation, acronym or other type. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a short form is treated as a separate concept. Consider the specific example of ‘API’ and ‘application programming interface’. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the relation between abbreviated and expanded designations. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. verify search expands the abbreviation without rewriting it. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

16. Source provenance

What this mechanism controls. Source provenance concerns where a definition, term or policy decision came from. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, an exporter discards bibliographic evidence. Consider the specific example of an official manual definition blended with a marketing blog term. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve a resolvable citation or stable evidence locator. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. trace the approval back to the source artifact. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

17. Responsibility

What this mechanism controls. Responsibility concerns who reviewed and approved a record. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a tool account appears as the linguistic decision-maker. Consider the specific example of a batch import falsely attributed to an approver. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve named editorial responsibility and review scope. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. compare audit events with the underlying decision record. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

18. Transaction history

What this mechanism controls. Transaction history concerns when a record changed and why. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, modified dates overwrite original creation evidence. Consider the specific example of a deprecated label reintroduced after a silent migration. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve change type, timestamp, rationale and previous state. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. replay the history and reproduce the current decision. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

19. Cross-reference target

What this mechanism controls. Cross-reference target concerns a relation to another concept or external authority. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, the link points to a discarded local identifier. Consider the specific example of a synonym link converted into a hierarchy relation. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve target identifier, relation meaning and link resolution. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. validate each target and inspect its relationship type. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

20. External reference

What this mechanism controls. External reference concerns a link to a document or outside terminology registry. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a fragile URL is presented as sufficient provenance. Consider the specific example of a vendor manual is replaced but the old wording remains. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve source identity even if its display address changes. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. open the reference and compare edition or revision. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

21. Customer subset

What this mechanism controls. Customer subset concerns the audience or client authorized to use a term. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, client-specific approvals bleed into global terminology. Consider the specific example of a product line requires a distinct branded phrase. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the subset rule and its parent concept relationship. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test both customer views and the global view. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

22. Project subset

What this mechanism controls. Project subset concerns a controlled release or deliverable scope. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, temporary project choices become permanent universal defaults. Consider the specific example of a campaign variation imported as official product terminology. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve scope dates, owner and explicit expiration policy. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. check the same term in a second unrelated project. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

23. Terminological status mapping

What this mechanism controls. Terminological status mapping concerns the relation between sender and receiver vocabularies of status. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a four-level model is compressed to approved or unapproved. Consider the specific example of deprecated and superseded collapsed into a single neutral state. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the precise source status and declared target conversion. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. publish a mapping table and enumerate irreversible losses. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

24. Dialect declaration

What this mechanism controls. Dialect declaration concerns the precise TBX dialect the file claims to implement. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a consumer assumes every TBX file accepts the same fields. Consider the specific example of a Basic-only data category added to a Min export. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve a named dialect and corresponding schema package. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. validate under the declared dialect rather than a permissive parser. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

25. TBX-Core boundary

What this mechanism controls. TBX-Core boundary concerns the shared structural base for compliant dialects. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, generic structure is mistaken for full domain-specific compliance. Consider the specific example of a structurally legal container holding an unpermitted category. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the distinction between core model and allowed extensions. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. separate core validation from dialect-specific validation. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

26. Data-category placement

What this mechanism controls. Data-category placement concerns which metadata belongs at concept, language or term level. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a term-level property is attached to an entire concept. Consider the specific example of part of speech mistakenly attached to all multilingual equivalents. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the level permitted by that category and chosen dialect. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. check level-specific rules using deliberately invalid examples. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

27. DCA and DCT representation

What this mechanism controls. DCA and DCT representation concerns the two XML styles of representing data categories. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a converter rewrites tags without respecting namespace meaning. Consider the specific example of an attribute-valued status lost during representation conversion. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve equivalent category semantics under the chosen style. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. compare the logical data model, not just element names. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

28. XML namespace

What this mechanism controls. XML namespace concerns the namespace distinguishing vocabulary elements and categories. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, an XML file parses but vocabulary elements are semantically misidentified. Consider the specific example of a default namespace changes after combining files. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve namespace-qualified names and resolver configuration. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. validate qualified names and imported data-category recognition. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

29. Character encoding

What this mechanism controls. Character encoding concerns the exact characters in term forms and notes. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, normalization or encoding damages accents or non-Latin text. Consider the specific example of a diacritic removed from a medicine name. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve Unicode text with tested round-trip behaviour. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. compare codepoints, visible rendering and searching. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

30. XML escaping

What this mechanism controls. XML escaping concerns literal punctuation versus XML syntax. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, ampersands or angle brackets break parsing or alter text. Consider the specific example of ‘R&D’ exported as malformed XML. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve correctly serialized text values. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. reparse and compare decoded text with the source. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

31. Language-tag discipline

What this mechanism controls. Language-tag discipline concerns the intended linguistic variety of each term and note. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, invalid or overly broad tags disguise meaning. Consider the specific example of Simplified and Traditional Chinese variants mixed. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve BCP 47 style language identification in XML language fields. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. validate tags and run separate locale views. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

32. Missing translation state

What this mechanism controls. Missing translation state concerns absence of target term versus negative or rejected candidate. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a blank is interpreted as an approved empty term. Consider the specific example of no safe equivalent exists for an administrative label. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve explicit state information outside placeholder translations. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. require review instead of inventing a translation. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

33. Multiple valid synonyms

What this mechanism controls. Multiple valid synonyms concerns several forms that are not necessarily interchangeable. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, an importer flattens nuance into an undifferentiated list. Consider the specific example of ‘meter’ and ‘metre’ with audience-specific preferences. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve per-term contexts, status and usage limits. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test sentence substitution in both recipient varieties. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

34. Homographs

What this mechanism controls. Homographs concerns identical spelling for different concepts. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, records keyed by term text are merged. Consider the specific example of ‘seal’ as security closure versus marine animal. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve separate concept identities and definitions. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. import two same-spelled labels and confirm two records. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

35. False friends

What this mechanism controls. False friends concerns similar-looking designations with different meanings. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, machine translation assumes orthographic similarity proves equivalence. Consider the specific example of ‘actual’ and an apparent cognate with different common meaning. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve domain evidence and bilingual reviewer confirmation. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. ask the reviewer to generate a disambiguating sentence. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

36. Numerical expression

What this mechanism controls. Numerical expression concerns quantities, dates and measurements inside context fields. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, locale formatting changes the represented value. Consider the specific example of 0.5 transformed into 5 during a comma conversion. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve the underlying number, unit and applicability condition. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. parse the quantity before and after text localization. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

37. Definition revision

What this mechanism controls. Definition revision concerns how a revised definition changes term validity. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, text is edited without reconsidering term approval. Consider the specific example of a standard narrows a technical class after update. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve revision identifier and an impact analysis of dependent terms. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. re-evaluate preferred status against the new boundary. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

38. Concept split

What this mechanism controls. Concept split concerns one entry becoming two distinct concepts. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, old terms are copied into both entries without evidence. Consider the specific example of ‘battery’ split by product chemistry classification. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve migration map from old concept ID to new IDs. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. check every old reference against both new meanings. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

39. Concept merge

What this mechanism controls. Concept merge concerns multiple records consolidated when they truly denote one concept. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, near-synonyms are merged because their labels resemble each other. Consider the specific example of two project-specific labels appear to overlap but differ in safety scope. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve evidence of shared defining characteristics. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. run counterexamples before approving the merge. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

40. Import loss reporting

What this mechanism controls. Import loss reporting concerns what the receiver discards or transforms. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, software reports success despite silent field loss. Consider the specific example of source notes vanish while term strings still appear. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve machine-readable reconciliation by field and record. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. produce counts and samples of every difference. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

41. Round-trip comparison

What this mechanism controls. Round-trip comparison concerns equivalence after export-import-export. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, formatting differences obscure semantic loss. Consider the specific example of IDs remapped but relationships become dangling. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve normalized logical model comparison with approved exceptions. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. compare graphs and statuses instead of raw byte equality. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

42. Termbase search

What this mechanism controls. Termbase search concerns how readers and tools retrieve correct records. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, stemming and fuzzy search return wrong concepts. Consider the specific example of medical search finds a deprecated everyday homograph. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve search behaviour aware of subject, status and language. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test exact, fuzzy and filtered searches separately. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

43. CAT-tool suggestion

What this mechanism controls. CAT-tool suggestion concerns the action produced by a terminology match. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a matching label appears in a translation suggestion without scope. Consider the specific example of an admitted form overrides a preferred medical device label. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve suggestion priority, prohibited-term rules and context. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test actual editing interface suggestions. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

44. Machine translation integration

What this mechanism controls. Machine translation integration concerns the way approved terminology influences generated output. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, a preference becomes an unconditional substitution. Consider the specific example of a term used correctly in prose but wrongly in a safety warning. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve context-aware constraints and human review. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. evaluate multiple contexts, not a single isolated term. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

45. Data access policy

What this mechanism controls. Data access policy concerns who may view internal terminology or citations. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, an export leaks confidential customer or project notes. Consider the specific example of a private product codename included in public TBX. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve least-privilege handling with designated disclosure fields. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. test exports under each intended permission. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

46. Privacy and personal data

What this mechanism controls. Privacy and personal data concerns sensitive names embedded in citations or reviewer notes. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, records unintentionally expose individuals. Consider the specific example of a clinical context sample includes a patient name. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve data minimization with legitimate retention purposes. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. scan contexts and attribution fields before release. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

47. Governance cadence

What this mechanism controls. Governance cadence concerns how new terms and revisions are reviewed over time. A reliable TBX exchange carries the relevant information across the boundary between people, machines and languages, rather than asking the receiving system to infer it from a display word. This matters whenever a termbase is used to make choices, not merely to display a list: a reviewer may judge a candidate, a writer may accept a suggestion, and a publishing process may enforce a status automatically. The reader should be able to reconstruct what the record author intended and which evidence makes that intention defensible.

The hidden failure. In practice, an immutable import is mistaken for live terminology governance. Consider the specific example of a newly prohibited designation continues to appear in old exports. A superficial verification might report a successful export because the text and file name survived. That is not an adequate acceptance test. The relevant claim is whether the receiving record still denotes the same concept, carries the same limitations and causes the same safe editorial behaviour. Before approving the exchange, identify which field holds this meaning, what the chosen dialect permits and which information the receiving tool is capable of representing. An unrecognized field is a loss to investigate, not a reason to pretend the meaning was never present.

A repeatable repair. Preserve review dates, owners, change controls and sunset policies. First inspect the source record in its original application and write a plain-language statement of the intended distinction. Then look at the actual exported XML and its declared dialect, checking whether the information is at concept, language or term level as appropriate. Import a small controlled sample rather than the entire production collection. Compare what is displayed, searchable and enforced. Where there is no lossless receiver mapping, preserve the original value in an accompanying audit record and ask the responsible terminology owner whether the transformation is acceptable. The fact that a conversion script can perform a change does not establish authority to make the change.

Diagnostic and transfer. trace downstream versions and confirm correction propagation. Use a positive example that should remain valid and a boundary example that should be rejected or escalated. Record the result, application version, source and target dialect, language tag, and reviewer. This is particularly important when a team later reuses the termbase for a different product: a choice that was safe for one interface or domain may not be safe as a universal language rule. The aim is a transferable explanation for each decision, not a fragile collection of unexplained exceptions.

5. Worked examples and diagnostics: a cross-domain casebook

Every case that follows is hypothetical and constructed for teaching. The cases are not excerpts from real patient records, regulatory submissions, contracts or private termbases. Each tests a different failure that can occur when concepts and terms move across systems. The format is deliberately operational: observe the symptom, identify what was actually asserted, perform a repair without inventing equivalence, and run a diagnostic that could reveal a regression.

Diagnostic 01 — Medical device instructions: ‘single use’ versus ‘one session’

Situation. In a fictional medical device instructions termbase, editors encounter ‘single use’ versus ‘one session’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because reuse prohibition is softened into a vague duration. A simple glossary export would not reliably explain this conflict. The exchange must preserve approved risk-control wording, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible clinical safety reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. try a scenario involving a device used twice in one appointment. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 02 — Hospital laboratory: ‘reference interval’ versus ‘normal value’

Situation. In a fictional hospital laboratory termbase, editors encounter ‘reference interval’ versus ‘normal value’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because population reference is converted into a personal clinical verdict. A simple glossary export would not reliably explain this conflict. The exchange must preserve laboratory concept definition, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible qualified laboratory terminologist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a borderline result with a clinical note. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 03 — Medication label: ‘once daily’ versus ‘every 24 hours’

Situation. In a fictional medication label termbase, editors encounter ‘once daily’ versus ‘every 24 hours’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because schedule becomes a rigid interval not stated by the source. A simple glossary export would not reliably explain this conflict. The exchange must preserve dose-frequency concept, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible medical safety approver should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare noon dosing and next-morning dosing. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 04 — Medical imaging: ‘contrast enhanced’ versus ‘contrast administered’

Situation. In a fictional medical imaging termbase, editors encounter ‘contrast enhanced’ versus ‘contrast administered’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because examination label is treated as proof a drug reached the patient. A simple glossary export would not reliably explain this conflict. The exchange must preserve exam metadata meaning, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible radiology reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. inspect a cancelled administration record. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 05 — Public health form: ‘suspected case’ versus ‘confirmed case’

Situation. In a fictional public health form termbase, editors encounter ‘suspected case’ versus ‘confirmed case’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because uncertainty disappears from the translated designation. A simple glossary export would not reliably explain this conflict. The exchange must preserve case-classification status, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible epidemiology specialist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an unconfirmed result in a report. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 06 — Clinical study: ‘adverse event’ versus ‘adverse reaction’

Situation. In a fictional clinical study termbase, editors encounter ‘adverse event’ versus ‘adverse reaction’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because causal implication is introduced where none was established. A simple glossary export would not reliably explain this conflict. The exchange must preserve causality distinction, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible study safety reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare an incident unrelated to treatment. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 07 — Financial reporting: ‘revenue’ versus ‘cash received’

Situation. In a fictional financial reporting termbase, editors encounter ‘revenue’ versus ‘cash received’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because accrual concept is replaced by cash movement. A simple glossary export would not reliably explain this conflict. The exchange must preserve accounting definition, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible financial reporting reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an unpaid invoice recognized as revenue. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 08 — Invoice processing: ‘credit note’ versus ‘refund’

Situation. In a fictional invoice processing termbase, editors encounter ‘credit note’ versus ‘refund’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because accounting adjustment is promised as cash payment. A simple glossary export would not reliably explain this conflict. The exchange must preserve transaction type, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible accounts-payable reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. trace whether money actually moved. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 09 — Insurance policy: ‘exclusion’ versus ‘limitation’

Situation. In a fictional insurance policy termbase, editors encounter ‘exclusion’ versus ‘limitation’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because different claim consequences are collapsed. A simple glossary export would not reliably explain this conflict. The exchange must preserve policy clause meaning, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible insurance terminology owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a claim falling just inside the boundary. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 10 — Investment document: ‘principal’ versus ‘capital gain’

Situation. In a fictional investment document termbase, editors encounter ‘principal’ versus ‘capital gain’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because the original amount is conflated with return. A simple glossary export would not reliably explain this conflict. The exchange must preserve investment concept scope, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible financial editor should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare a redemption below initial investment. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 11 — Software permissions: ‘view’ versus ‘edit’

Situation. In a fictional software permissions termbase, editors encounter ‘view’ versus ‘edit’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because read-only permission becomes a write grant. A simple glossary export would not reliably explain this conflict. The exchange must preserve access-control role label, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible security reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. try a user without update privileges. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 12 — Authentication interface: ‘verify account’ versus ‘sign in’

Situation. In a fictional authentication interface termbase, editors encounter ‘verify account’ versus ‘sign in’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because identity confirmation is confused with session creation. A simple glossary export would not reliably explain this conflict. The exchange must preserve action and user-state distinction, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible product security owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test when the user has no active session. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 13 — Error notification: ‘request failed’ versus ‘service unavailable’

Situation. In a fictional error notification termbase, editors encounter ‘request failed’ versus ‘service unavailable’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because a general failure is framed as verified outage. A simple glossary export would not reliably explain this conflict. The exchange must preserve error taxonomy and status code, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible API product owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare 400-series and 500-series responses. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 14 — API documentation: ‘rate limit’ versus ‘usage quota’

Situation. In a fictional api documentation termbase, editors encounter ‘rate limit’ versus ‘usage quota’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because time-window throttling is confused with subscription entitlement. A simple glossary export would not reliably explain this conflict. The exchange must preserve limit type and reset semantics, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible API maintainer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a short burst followed by recovery. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 15 — Data pipeline: ‘source event time’ versus ‘ingestion time’

Situation. In a fictional data pipeline termbase, editors encounter ‘source event time’ versus ‘ingestion time’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because chronology is reordered after a late event. A simple glossary export would not reliably explain this conflict. The exchange must preserve timestamp meaning, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible data engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. replay an event arriving days late. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 16 — Cloud storage: ‘archive’ versus ‘delete’

Situation. In a fictional cloud storage termbase, editors encounter ‘archive’ versus ‘delete’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because a reversible state is rendered as irreversible removal. A simple glossary export would not reliably explain this conflict. The exchange must preserve retention state and operation, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible platform owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check retrieval after the operation. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 17 — Cybersecurity: ‘observed’ versus ‘attributed’

Situation. In a fictional cybersecurity termbase, editors encounter ‘observed’ versus ‘attributed’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because a detection claim becomes a claim about responsibility. A simple glossary export would not reliably explain this conflict. The exchange must preserve confidence and evidence boundary, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible threat analyst should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an indicator shared by two actors. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 18 — Threat intelligence: ‘indicator’ versus ‘incident’

Situation. In a fictional threat intelligence termbase, editors encounter ‘indicator’ versus ‘incident’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because a signal is made to imply an occurred attack. A simple glossary export would not reliably explain this conflict. The exchange must preserve event-evidence separation, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible intelligence reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare an untriggered rule with confirmed breach. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 19 — Privacy notice: ‘may share’ versus ‘will share’

Situation. In a fictional privacy notice termbase, editors encounter ‘may share’ versus ‘will share’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because a possibility is rewritten as certainty. A simple glossary export would not reliably explain this conflict. The exchange must preserve modal force of policy statement, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible privacy counsel should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare a conditional consent path. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 20 — Consent form: ‘withdraw consent’ versus ‘erase data’

Situation. In a fictional consent form termbase, editors encounter ‘withdraw consent’ versus ‘erase data’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because future processing restriction becomes an erasure promise. A simple glossary export would not reliably explain this conflict. The exchange must preserve legal action distinction, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible privacy governance owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test retained records under a lawful basis. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 21 — Manufacturing: ‘batch’ versus ‘lot’

Situation. In a fictional manufacturing termbase, editors encounter ‘batch’ versus ‘lot’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because distinct traceability groupings are merged. A simple glossary export would not reliably explain this conflict. The exchange must preserve production identifier mapping, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible quality manager should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. trace one shipment containing multiple batches. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 22 — Manufacturing quality: ‘nonconformity’ versus ‘defect’

Situation. In a fictional manufacturing quality termbase, editors encounter ‘nonconformity’ versus ‘defect’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because failure against a requirement is treated as physical flaw. A simple glossary export would not reliably explain this conflict. The exchange must preserve quality taxonomy, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible quality engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a missing signature with sound product. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 23 — Machinery safety: ‘isolate power’ versus ‘switch off’

Situation. In a fictional machinery safety termbase, editors encounter ‘isolate power’ versus ‘switch off’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because electrical isolation is reduced to ordinary shutdown. A simple glossary export would not reliably explain this conflict. The exchange must preserve safety procedure boundary, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible safety engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a system with residual energy. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 24 — Maintenance record: ‘preventive’ versus ‘corrective’

Situation. In a fictional maintenance record termbase, editors encounter ‘preventive’ versus ‘corrective’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because work category and cause become inverted. A simple glossary export would not reliably explain this conflict. The exchange must preserve maintenance task classification, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible asset manager should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare scheduled inspection with fault repair. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 25 — Logistics tracking: ‘delivered to carrier’ versus ‘delivered to recipient’

Situation. In a fictional logistics tracking termbase, editors encounter ‘delivered to carrier’ versus ‘delivered to recipient’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because custody transfer becomes final delivery. A simple glossary export would not reliably explain this conflict. The exchange must preserve event location and party, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible logistics owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. trace a package held at a hub. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 26 — Supply-chain traceability: ‘commissioned’ versus ‘shipped’

Situation. In a fictional supply-chain traceability termbase, editors encounter ‘commissioned’ versus ‘shipped’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because creation of an item identity is confused with physical movement. A simple glossary export would not reliably explain this conflict. The exchange must preserve business-step meaning, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible traceability specialist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an item never leaving the factory. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 27 — Retail pricing: ‘before tax’ versus ‘tax included’

Situation. In a fictional retail pricing termbase, editors encounter ‘before tax’ versus ‘tax included’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because display label changes charge interpretation. A simple glossary export would not reliably explain this conflict. The exchange must preserve price basis field, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible commerce policy owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare a final checkout amount. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 28 — Subscription commerce: ‘pause’ versus ‘cancel’

Situation. In a fictional subscription commerce termbase, editors encounter ‘pause’ versus ‘cancel’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because temporary suspension is presented as termination. A simple glossary export would not reliably explain this conflict. The exchange must preserve contract state semantics, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible subscription owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. reactivate and inspect billing state. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 29 — Return policy: ‘exchange’ versus ‘refund’

Situation. In a fictional return policy termbase, editors encounter ‘exchange’ versus ‘refund’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because replacement is presented as money-back entitlement. A simple glossary export would not reliably explain this conflict. The exchange must preserve remedy conditions, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible consumer policy reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a defective item with exchange-only terms. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 30 — Legal drafting: ‘shall’ versus ‘may’

Situation. In a fictional legal drafting termbase, editors encounter ‘shall’ versus ‘may’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because obligation is turned into permission. A simple glossary export would not reliably explain this conflict. The exchange must preserve normative force, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible legal language reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare compliance consequences. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 31 — Contract management: ‘effective date’ versus ‘signing date’

Situation. In a fictional contract management termbase, editors encounter ‘effective date’ versus ‘signing date’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because a legal start condition is shifted in time. A simple glossary export would not reliably explain this conflict. The exchange must preserve date role, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible contract owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a delayed commencement clause. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 32 — Employment policy: ‘eligible’ versus ‘entitled’

Situation. In a fictional employment policy termbase, editors encounter ‘eligible’ versus ‘entitled’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because conditional qualification becomes guaranteed benefit. A simple glossary export would not reliably explain this conflict. The exchange must preserve eligibility criteria, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible HR policy reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test someone who meets only one criterion. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 33 — Recruitment form: ‘preferred qualification’ versus ‘required qualification’

Situation. In a fictional recruitment form termbase, editors encounter ‘preferred qualification’ versus ‘required qualification’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because selection advice becomes exclusion criterion. A simple glossary export would not reliably explain this conflict. The exchange must preserve requirement status, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible hiring compliance owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a candidate lacking the preference. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 34 — School assessment: ‘formative’ versus ‘summative’

Situation. In a fictional school assessment termbase, editors encounter ‘formative’ versus ‘summative’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because feedback-oriented activity is mistaken for final grading. A simple glossary export would not reliably explain this conflict. The exchange must preserve assessment purpose, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible education measurement reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test practice that does not affect final score. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 35 — Learning platform: ‘attempt’ versus ‘submission’

Situation. In a fictional learning platform termbase, editors encounter ‘attempt’ versus ‘submission’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because each start is counted as delivered answer. A simple glossary export would not reliably explain this conflict. The exchange must preserve workflow state, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible assessment engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check an abandoned quiz. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 36 — Research methods: ‘correlation’ versus ‘causal effect’

Situation. In a fictional research methods termbase, editors encounter ‘correlation’ versus ‘causal effect’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because statistical association is turned into mechanism. A simple glossary export would not reliably explain this conflict. The exchange must preserve study claim strength, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible research methodologist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a confounded observational study. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 37 — Scientific measurement: ‘accuracy’ versus ‘precision’

Situation. In a fictional scientific measurement termbase, editors encounter ‘accuracy’ versus ‘precision’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because bias and repeatability are interchanged. A simple glossary export would not reliably explain this conflict. The exchange must preserve measurement-property concept, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible metrology reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare clustered inaccurate readings. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 38 — Survey design: ‘response rate’ versus ‘completion rate’

Situation. In a fictional survey design termbase, editors encounter ‘response rate’ versus ‘completion rate’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because different denominators are conflated. A simple glossary export would not reliably explain this conflict. The exchange must preserve population and denominator, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible survey methodologist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. calculate both on a sample with partial forms. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 39 — Environment report: ’emissions’ versus ‘concentrations’

Situation. In a fictional environment report termbase, editors encounter ’emissions’ versus ‘concentrations’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because quantity released is confused with measured ambient level. A simple glossary export would not reliably explain this conflict. The exchange must preserve physical quantity and unit, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible environmental scientist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check a source with low ambient concentration. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 40 — Energy reporting: ‘power’ versus ‘energy’

Situation. In a fictional energy reporting termbase, editors encounter ‘power’ versus ‘energy’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because kilowatts and kilowatt-hours are treated as synonyms. A simple glossary export would not reliably explain this conflict. The exchange must preserve dimensional distinction, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible energy engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compute consumption over two hours. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 41 — Water testing: ‘detection limit’ versus ‘quantitation limit’

Situation. In a fictional water testing termbase, editors encounter ‘detection limit’ versus ‘quantitation limit’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because ability to detect is presented as reliable measurement. A simple glossary export would not reliably explain this conflict. The exchange must preserve analytical threshold, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible laboratory analyst should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a trace signal below quantitation. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 42 — Transport timetable: ‘departure’ versus ‘boarding’

Situation. In a fictional transport timetable termbase, editors encounter ‘departure’ versus ‘boarding’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because passenger action timing becomes vehicle movement. A simple glossary export would not reliably explain this conflict. The exchange must preserve event type, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible operations planner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check a late boarding scenario. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 43 — Aviation operations: ‘estimated’ versus ‘actual’

Situation. In a fictional aviation operations termbase, editors encounter ‘estimated’ versus ‘actual’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because forecast event is converted into verified event. A simple glossary export would not reliably explain this conflict. The exchange must preserve temporal certainty, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible flight operations reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. inspect a diversion after estimate. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 44 — Maritime logistics: ‘draft’ versus ‘displacement’

Situation. In a fictional maritime logistics termbase, editors encounter ‘draft’ versus ‘displacement’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because vessel depth is confused with mass of displaced water. A simple glossary export would not reliably explain this conflict. The exchange must preserve hydrostatic concept, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible marine engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare ships of different beam. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 45 — Rail operations: ‘occupied’ versus ‘reserved’

Situation. In a fictional rail operations termbase, editors encounter ‘occupied’ versus ‘reserved’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because physical train presence is confused with future allocation. A simple glossary export would not reliably explain this conflict. The exchange must preserve signalling state, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible rail safety engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a reserved but empty section. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 46 — Construction BIM: ‘approved’ versus ‘issued for review’

Situation. In a fictional construction bim termbase, editors encounter ‘approved’ versus ‘issued for review’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because document workflow status escalates without authorization. A simple glossary export would not reliably explain this conflict. The exchange must preserve revision and release state, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible design coordinator should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare preliminary and construction issue. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 47 — Civil engineering: ‘elevation’ versus ‘height above ground’

Situation. In a fictional civil engineering termbase, editors encounter ‘elevation’ versus ‘height above ground’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because reference datum is lost. A simple glossary export would not reliably explain this conflict. The exchange must preserve vertical datum, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible surveyor should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare two local benchmarks. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 48 — Geospatial metadata: ‘position accuracy’ versus ‘precision’

Situation. In a fictional geospatial metadata termbase, editors encounter ‘position accuracy’ versus ‘precision’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because quality measures are represented as interchangeable. A simple glossary export would not reliably explain this conflict. The exchange must preserve coordinate-quality dimension, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible GIS reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an accurate but noisy dataset. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 49 — Weather bulletin: ‘watch’ versus ‘warning’

Situation. In a fictional weather bulletin termbase, editors encounter ‘watch’ versus ‘warning’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because different certainty or urgency classes are merged. A simple glossary export would not reliably explain this conflict. The exchange must preserve hazard bulletin classification, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible meteorology reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a potential event not yet occurring. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 50 — Disaster response: ‘evacuation advisory’ versus ‘order’

Situation. In a fictional disaster response termbase, editors encounter ‘evacuation advisory’ versus ‘order’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because recommendation becomes binding instruction. A simple glossary export would not reliably explain this conflict. The exchange must preserve authority and urgency, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible emergency communications owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a voluntary departure notice. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 51 — Telecommunications: ‘latency’ versus ‘throughput’

Situation. In a fictional telecommunications termbase, editors encounter ‘latency’ versus ‘throughput’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because delay is mistaken for carried data rate. A simple glossary export would not reliably explain this conflict. The exchange must preserve performance metric, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible network engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a slow response on high-bandwidth link. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 52 — Electronics: ‘ground’ versus ‘neutral’

Situation. In a fictional electronics termbase, editors encounter ‘ground’ versus ‘neutral’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because different electrical roles are merged. A simple glossary export would not reliably explain this conflict. The exchange must preserve circuit function and safety, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible electrical engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an unbalanced load. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 53 — Battery engineering: ‘state of charge’ versus ‘state of health’

Situation. In a fictional battery engineering termbase, editors encounter ‘state of charge’ versus ‘state of health’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because available charge level is confused with long-term degradation. A simple glossary export would not reliably explain this conflict. The exchange must preserve battery metric, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible battery systems engineer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare a full but aged battery. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 54 — Robotics: ’emergency stop’ versus ‘protective stop’

Situation. In a fictional robotics termbase, editors encounter ’emergency stop’ versus ‘protective stop’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because distinct safety and recovery behaviours are collapsed. A simple glossary export would not reliably explain this conflict. The exchange must preserve machine safety state, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible robotics safety reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test restart authorization. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 55 — Chemistry: ‘flammable’ versus ‘combustible’

Situation. In a fictional chemistry termbase, editors encounter ‘flammable’ versus ‘combustible’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because controlled classification differences are treated as style. A simple glossary export would not reliably explain this conflict. The exchange must preserve hazard classification basis, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible chemical safety reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare product SDS category. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 56 — Food labelling: ‘use by’ versus ‘best before’

Situation. In a fictional food labelling termbase, editors encounter ‘use by’ versus ‘best before’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because safety deadline and quality date are equated. A simple glossary export would not reliably explain this conflict. The exchange must preserve date purpose, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible food compliance reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test stored food after its quality date. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 57 — Agricultural data: ‘yield’ versus ‘production volume’

Situation. In a fictional agricultural data termbase, editors encounter ‘yield’ versus ‘production volume’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because per-area ratio becomes absolute amount. A simple glossary export would not reliably explain this conflict. The exchange must preserve measurement denominator, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible agronomy reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare farms with different area. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 58 — Biology taxonomy: ‘species’ versus ‘subspecies’

Situation. In a fictional biology taxonomy termbase, editors encounter ‘species’ versus ‘subspecies’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because rank relationship is flattened. A simple glossary export would not reliably explain this conflict. The exchange must preserve taxonomic rank, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible taxonomy specialist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check a trinomial designation. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 59 — Cultural heritage: ‘attributed to’ versus ‘made by’

Situation. In a fictional cultural heritage termbase, editors encounter ‘attributed to’ versus ‘made by’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because scholarly uncertainty becomes proven authorship. A simple glossary export would not reliably explain this conflict. The exchange must preserve evidence status, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible curatorial reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test disputed maker identity. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 60 — Archival catalog: ‘creator’ versus ‘owner’

Situation. In a fictional archival catalog termbase, editors encounter ‘creator’ versus ‘owner’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because record provenance is confused with custody. A simple glossary export would not reliably explain this conflict. The exchange must preserve archival role, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible archivist should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. trace a donated collection. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 61 — Music catalog: ‘arranger’ versus ‘composer’

Situation. In a fictional music catalog termbase, editors encounter ‘arranger’ versus ‘composer’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because different creative contributions are merged. A simple glossary export would not reliably explain this conflict. The exchange must preserve contribution role, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible music rights editor should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a new arrangement of old work. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 62 — Audiovisual: ‘caption’ versus ‘subtitle’

Situation. In a fictional audiovisual termbase, editors encounter ‘caption’ versus ‘subtitle’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because sound-effect accessibility content is lost. A simple glossary export would not reliably explain this conflict. The exchange must preserve track purpose, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible accessibility reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a dialogue-free sound event. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 63 — Publishing: ‘edition’ versus ‘printing’

Situation. In a fictional publishing termbase, editors encounter ‘edition’ versus ‘printing’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because content revision is confused with manufacture run. A simple glossary export would not reliably explain this conflict. The exchange must preserve bibliographic relationship, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible metadata editor should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. compare reprint with revised text. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 64 — Public procurement: ‘bidder’ versus ‘awardee’

Situation. In a fictional public procurement termbase, editors encounter ‘bidder’ versus ‘awardee’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because application status becomes confirmed contract. A simple glossary export would not reliably explain this conflict. The exchange must preserve award stage, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible procurement reviewer should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an unsuccessful valid bid. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 65 — Voting records: ‘abstained’ versus ‘absent’

Situation. In a fictional voting records termbase, editors encounter ‘abstained’ versus ‘absent’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because participation is misreported. A simple glossary export would not reliably explain this conflict. The exchange must preserve participation status, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible records clerk should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a present member who did not vote. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 66 — Charity reporting: ‘restricted fund’ versus ‘designated fund’

Situation. In a fictional charity reporting termbase, editors encounter ‘restricted fund’ versus ‘designated fund’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because external restriction is confused with internal board allocation. A simple glossary export would not reliably explain this conflict. The exchange must preserve fund restriction source, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible nonprofit accountant should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test reallocation by the board. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 67 — Public service: ‘application received’ versus ‘approved’

Situation. In a fictional public service termbase, editors encounter ‘application received’ versus ‘approved’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because administrative intake is mistaken for successful decision. A simple glossary export would not reliably explain this conflict. The exchange must preserve process milestone, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible service owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test an incomplete application. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 68 — Humanitarian logistics: ‘dispatched’ versus ‘distributed’

Situation. In a fictional humanitarian logistics termbase, editors encounter ‘dispatched’ versus ‘distributed’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because goods sent from a warehouse appear delivered to beneficiaries. A simple glossary export would not reliably explain this conflict. The exchange must preserve custody state, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible relief coordinator should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check a shipment awaiting last-mile transport. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 69 — Software localization: ‘save draft’ versus ‘publish’

Situation. In a fictional software localization termbase, editors encounter ‘save draft’ versus ‘publish’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because private state becomes public release. A simple glossary export would not reliably explain this conflict. The exchange must preserve workflow action, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible content product owner should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test the item visibility immediately after action. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 70 — Terminology itself: ‘term’ versus ‘concept’

Situation. In a fictional terminology itself termbase, editors encounter ‘term’ versus ‘concept’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because word form is used as if it were meaning identity. A simple glossary export would not reliably explain this conflict. The exchange must preserve concept-oriented structure, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible terminology architect should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a homonym with two definitions. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 71 — Localization governance: ‘obsolete’ versus ‘deprecated’

Situation. In a fictional localization governance termbase, editors encounter ‘obsolete’ versus ‘deprecated’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because policy distinctions collapse without mapping. A simple glossary export would not reliably explain this conflict. The exchange must preserve source status semantics, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible termbase administrator should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. check an old label that is still searchable. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

Diagnostic 72 — Technical standards: ‘shall’ versus ‘should’

Situation. In a fictional technical standards termbase, editors encounter ‘shall’ versus ‘should’. The two expressions may appear close in everyday reading, but the controlled record must establish whether they represent one concept, distinct concepts or a constrained relation. An incoming TBX export triggers an alarm because requirement becomes recommendation. A simple glossary export would not reliably explain this conflict. The exchange must preserve normative verbal force, together with a definition, any applicable status and enough context for a reviewer to decide what is actually allowed. A term that reads fluently can still make the record wrong.

Diagnosis and repair. Separate the two meanings in a concept table before selecting target-language designations. Check whether the source used the term field as a surrogate identifier, whether language sections were copied, and whether a receiving application dropped the distinguishing note. The responsible standards editor should adjudicate the boundary instead of relying on a bilingual synonym list. Repair the record at its proper level; do not alter unrelated concepts to make a test pass. Preserve the original export and an explicit record of the change, so a later migration can explain why this entry differs from its source.

Proof. test a conformance clause. Repeat the test in the receiving termbase, an authoring search and the exported TBX result. Confirm that the decision remains consistent when someone searches for a disallowed alternative, selects a preferred term, or requests the same concept in another language. The pass condition is not universal word-for-word sameness: it is correct meaning, appropriate permission, reliable evidence and a reproducible review path. Record a counterexample as well as the successful example so the scope cannot drift silently.

6. Release validation: XML syntax is necessary, not sufficient

A validation pipeline can be implemented in layers. Layer one checks whether the file can be parsed as XML and preserves encoded characters. Layer two checks the chosen TBX version and dialect against the applicable schema set, including namespace and data-category placement. Layer three reconciles counts and identifiers. Layer four applies terminology policy tests: preferred status, forbidden forms, subject restrictions, definition coherence and evidence links. Layer five performs a real target-system import, then compares the resulting logical records with the source. Layer six checks human-facing use: search, autocomplete, warnings, copy generation and export. None of these layers can be substituted by a screenshot saying ‘import successful’.

For example, record three source counts: concepts, language sections, and term sections. Also record counts by language, status, subject and source type. After import, differences require explanation. A target that deduplicates literal forms may reduce its display count without losing conceptual meaning, but a target that merges separate concepts must not be approved just because it reports fewer duplicates. If identifiers are remapped, supply a reversible crosswalk. If a category has no corresponding field, document the precise loss and make a governance decision; do not hide it in an ‘other’ bucket that nobody reads.

Validation should include deliberate negative tests. Export a term with an ampersand, a concept with two homographic terms in different domains, a deprecated designation, an empty target-language section, a restricted citation and a reference to another concept. An application that handles only ordinary positive examples may fail exactly when the terminology is most important. Negative cases provide stronger evidence of robustness because they show where a system refuses unsafe guesses.

7. Transfer workshops for teams, developers and reviewers

Workshop 1: A 50-record pilot

Task. Build a pilot with one intentional example of each supported term status and every target language. Annotate expected behaviour before import, then inspect all fifty records by ID instead of spot-checking only the first few. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 2: The dual homograph

Task. Create two concept entries with the same English term but different definitions and subject fields. Assign two different target-language designations. Confirm the receiving application offers separate results rather than one merged entry. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 3: The missing-language test

Task. Include a concept with a reviewed source definition but no approved target-language term. Prove that the receiving platform represents the gap as missing work, rather than accepting an untranslated source string as approved terminology. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 4: The reversible status map

Task. Write a table of source statuses, receiving statuses and conditional conversions. Identify which pairs cannot be reversed exactly. Submit those ambiguities for governance sign-off before deploying a one-way converter. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 5: A language-tag comparison

Task. Add two genuine regional variants and one script distinction. Confirm that lookup, display and export keep all three distinctions. Do not conclude that a display flag or language name alone proves correct tagging. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 6: Source citation drill

Task. Select ten definitions with different documentary sources. Open every citation and verify the edition or timestamp, then compare source references after import and subsequent export. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 7: Round-trip equivalence

Task. Normalize XML whitespace, attribute ordering and harmless formatting differences. Compare logical concept IDs, term values, statuses, languages, references and notes separately. Fail the release on unexplained semantic changes. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 8: Controlled acronym practice

Task. Use an expanded form, acronym, abbreviated style and a similar-looking unrelated acronym. Ask whether the tool picks the right concept in running text without replacing a separate domain’s acronym. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 9: The forbidden-term exercise

Task. Mark a misleading phrase as disallowed for new publication but searchable for historical comprehension. Check authoring warnings, archive search and target-language report output without deleting useful historical evidence. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 10: An importer capability matrix

Task. Document which fields the receiving application supports, ignores, flattens or stores as attachments. Test the claims with real records and version numbers, not only a vendor feature list. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 11: A conflict-resolution meeting

Task. Present two reviewers with a plausible but inconsistent definition pair. Require a recorded concept decision, named approver, scope and effective date. Re-export after approval to ensure editorial reasoning survived. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 12: The terminology change notice

Task. Deprecate an old preferred label, introduce a reviewed replacement and deliver a change notice. Confirm that downstream authors receive the revision while historic citations remain traceable. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 13: A technical manual translation

Task. Apply the termbase to a paragraph containing three overlapping technical senses, a warning and an abbreviation. Record each suggestion used, each exception and the supporting concept entry. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 14: Machine translation boundaries

Task. Feed a model one phrase in three contexts with different concepts. Observe whether a generic term list forces the same translation in all contexts. Repair context and reviewer controls rather than demanding blind replacement. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 15: The spreadsheet conversion

Task. Import a two-column glossary with repeated labels. Add explicit concept identifiers and definitions before converting it to TBX. Note which relationships are genuinely unknown instead of inventing them. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 16: Version 2 to 3 migration

Task. Analyze an older TBX-style file and map its dialect and category representation to the chosen version 3 target. Validate before and after; save an exception report for constructs that have no direct counterpart. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 17: Multilingual terminology audit

Task. Pick one concept and compare definitions, grammatical categories, status values and contexts across every language section. Confirm conceptual alignment, not identical paragraph length. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 18: Public and private exports

Task. Prepare different export profiles for translators, authors and external vendors. Confirm that personal reviewer notes and confidential product references do not appear in the external package. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 19: Application upgrade rehearsal

Task. Re-run a frozen collection through the newest importing tool. Compare its warnings and semantic output to the previous version before switching production systems. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

Workshop 20: Final release simulation

Task. Ask a reviewer who did not build the converter to find one concept, inspect its alternatives, trace a decision, test a restricted term and export again. Their ability to explain the result is evidence of a maintainable interchange. Start from a written statement of what must remain true. Give participants a source copy, target application details, a declared dialect and a reviewer role. Require an explicit result rather than an impression that the file ‘looks good’.

Record and transfer. For this exercise, capture the source concept IDs, language tags, relevant categories, output counts, warnings and a before/after comparison. Ask someone to describe the business or linguistic consequence of a failure: would a writer choose the wrong term, would an importer collapse a concept, or would an auditor lose the evidence for a decision? Transfer the learned control into the team’s standard test fixture. When the next dialect or software version arrives, the same fixture should reveal whether the improvement changed semantic behaviour.

8. Frequently asked questions

1. What is TBX 3.0?

TBX version 3 is the third major generation of the TermBase eXchange framework associated with ISO 30042:2019. It helps exchange structured, concept-oriented terminology; it does not itself translate prose or guarantee the accuracy of term choices.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

2. Is TBX the same as a bilingual glossary?

No. A spreadsheet can be a starting source, but a dependable termbase may need concepts, definitions, languages, term status, grammatical information, source references and domain boundaries. A two-column file cannot convey all those properties automatically.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

3. Does .tbx guarantee compatibility?

No. The file extension alone does not disclose the dialect, category subset or importer capabilities. Check the declaration, schema and application support, then test a sample.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

4. What does ISO 30042:2019 specify?

It describes fundamental TBX concepts, the metamodel, XML representation styles, data categories and the approach for defining dialects. Its public page identifies TBX-Core as the dialect described in detail by the standard.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

5. Is a revision of ISO 30042 already published?

The ISO publication page still identifies ISO 30042:2019 as published and indicates revision activity. Check the official status when building a new project instead of assuming a proposed update is already an applicable edition.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

6. What is a TBX dialect?

A dialect is a defined set of structural and data-category choices and constraints within the TBX framework. Import and export agreements should name the chosen dialect explicitly.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

7. What is the distinction between concept and term?

The concept is the intended unit of knowledge; a term is a linguistic designation. One concept can have multiple terms and languages, and one visible spelling can denote different concepts.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

8. Where should a preferred-term status go?

It belongs with the term whose status is controlled, within the categories and placement permitted by the chosen dialect. Do not attach one synonym’s approval to all terms under a concept.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

9. Should forbidden terms be deleted?

Not necessarily. A forbidden or deprecated designation may need to remain searchable for historical material while being blocked or warned against in new writing. Preserve the exact policy distinction.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

10. Can a glossary have two preferred terms?

That depends on the organization’s concept model, language scope and term policy. Apparent duplicates may be regional or register-specific; do not make a global rule from a tool’s default display.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

11. Are language tags merely cosmetic?

No. They affect selection, search, rendering, linguistic rules and exchange. A wrongly tagged term can appear in the wrong language or regional list.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

12. How do I translate a definition?

First establish the concept and evidence. Write a clear target-language explanation that preserves distinguishing characteristics and exclusions, then have an appropriate domain reviewer compare it to the source.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

13. Can AI create target-language terms?

It can propose candidates. Domain accuracy, meaning boundaries, status approval and documented review are still needed, especially for safety-critical or regulated terminology.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

14. Can TBX store terminology governance?

The exact fields available depend on the chosen dialect and application. Useful governance information includes term status, evidence, responsibility, scope and revision history, but never assume a receiver supports every category.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

15. Is a successful XML parse enough?

No. It proves only syntax at that layer. Test dialect validation, information retention, concept correctness and behaviour in the actual receiving tool.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

16. Why did my importer remove fields?

Possible causes include unsupported categories, wrong dialect assumptions, incorrect namespace, invalid placement, mapping gaps or application restrictions. Compare field-level evidence rather than guessing.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

17. Should termbase XML be edited by hand?

Small fixes can be made carefully, but manual changes can introduce structure, identifier or reference errors. Use validation and a reviewed change trail, and keep the original export.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

18. How do I test a migration?

Count and classify source records, define a mapping, import a representative fixture, compare logical records, investigate differences and round-trip the result before full release.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

19. What is a round trip?

Exporting from a source, importing into a destination and exporting again to compare the information. Byte-for-byte identity may not be required, but semantic equivalence and any permitted losses must be demonstrated.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

20. Can two terms with the same spelling be merged?

Only when domain and definition evidence shows they denote the same concept. Identical spelling alone is not sufficient.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

21. What about translation memory?

Translation memory and terminology have complementary jobs. Segment history can supply useful context, but a segment pair is not automatically a validated concept record.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

22. What about XLIFF?

XLIFF exchanges localizable content and supporting metadata. TBX exchanges concept-oriented terminology. They may appear together in a localization pipeline but are not interchangeable.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

23. What does a source reference prove?

It provides a way to inspect where a claim came from. Source reliability, edition and relevance still need evaluation; a link alone is not proof that a term is correct.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

24. Can one source concept have no target term?

Yes. A missing approved designation is a legitimate state. Inventing a convenient equivalent can be worse than preserving the unresolved gap.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

25. How do we manage regional variation?

Represent actual differences in term usage with appropriate language and usage information supported by the dialect. Keep the underlying concept common when its meaning is genuinely shared.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

26. How should deprecated terms appear in search?

They can be displayed with an explicit warning and alternative, subject to governance policy. Do not silently remove the history or suggest them as current defaults.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

27. How do we protect confidential termbases?

Create separate export scopes, minimize sensitive notes, control recipients and test the actual exported package. An XML file can carry private information in contextual examples and citations.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

28. Do all CAT tools support the same TBX dialect?

No. Tools differ in supported dialects, categories, import behaviour and version handling. Test the product and version you intend to deploy.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

29. Who should approve the vocabulary mapping?

A terminology owner with relevant domain knowledge should approve conceptual distinctions and status decisions. Developers can implement the mapping but should not invent editorial policy.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

30. Should we translate identifiers?

Stable machine identifiers, controlled codes and references generally need preservation or explicit authorized remapping rather than natural-language translation. Verify actual field semantics first.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

31. How do we know a termbase is production ready?

It has a documented dialect and field map, structural and semantic checks, a verified target-system import, explained exceptions, reviewer approval and a reproducible rollback path.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

32. Where should a newcomer start?

Begin with a small concept-oriented glossary, add definitions, distinguish preferred and deprecated forms, choose a supported dialect, test the import and inspect one counterexample before scaling.

For a real project, convert this answer into a test with one accepted record and one deliberately difficult record. Specify who decides the concept boundary and what documentary evidence will settle an ambiguous translation. Keep the result close to the termbase rather than burying it in private discussion. If your receiving software cannot represent that decision, report the limitation explicitly and choose an approved alternative workflow. The meaningful objective is not merely to make a terminology file load; it is to preserve the reasons that trustworthy language choices can be made from it.

9. Quality-control worksheet and release decisions

A practical reviewer worksheet should contain: record identifier; source and target language tags; source and target dialects; concept definition; domain; approved designations; admitted forms; restricted or deprecated forms; source citation; content owner; importer product and version; validation result; mapping warnings; manual decision; decision date; and round-trip result. Maintain both a tabular view for triage and original structured evidence for reprocessing. The worksheet is a review instrument, not a substitute for the actual TBX file.

Severity can be determined by what a loss causes. Critical issues change a safety instruction, legal obligation, identity or controlled status. Major issues remove domain boundaries, source evidence or reliable search distinctions. Moderate issues affect usability or nonessential metadata but may still require repair before broad reuse. Minor issues are cosmetic transformations verified not to change meaning. Severity should be explicitly defined by the team’s risk profile; the examples here are practical categories, not assertions that ISO 5060 specifies these exact thresholds. ISO 5060:2024 discusses analytic evaluation of translation output using error types and penalty points, while workflow and corrective action also require separate controls.

Release should fail when unexplained concepts disappear, source and target counts cannot be reconciled, forbidden terms become approved, language tags are wrong, unresolved identifiers are silently reused, or protected notes escape into an unauthorized export. Release can proceed with documented, approved nonsemantic formatting changes, such as normalized XML whitespace, only after the project verifies that behaviour is equivalent. Treat every exception as test data for the next software change.

10. Research notes, standards status and authority

This article was researched against the public specification and guidance pages listed below on 1 November 2026. ISO 30042:2019 is the published TBX edition described by ISO; TBX Info explains version 3, dialects, tooling and validation. ISO 16642:2025 is the published terminology-markup framework, while ISO 704:2022 addresses concepts, definitions and designations. The W3C language-tag guide explains how language attributes identify linguistic content. ISO 5060:2024 is relevant to translation-output evaluation, but it should not be presented as a complete TBX import/export specification. This guide offers original teaching procedures, not a reproduction of restricted standards text or an official conformity certificate.

• ISO 30042:2019 — Management of terminology resources, TermBase eXchange
• TBX Info — overview, current dialects and validation resources
• TBX Info — TBX-Basic data categories
• ISO 16642:2025 — Terminological markup framework
• ISO 704:2022 — Terminology work, principles and methods
• ISO 5060:2024 — Evaluation of translation output
• W3C Internationalization — Language tags in HTML and XML

11. Next routes in the translation learning architecture

For the wider theory of moving meaning across languages, start with Master Art of Translation | The Complete System for Moving Meaning Between Languages. For how a learner develops distinctions between words, concepts and usage, continue to the protected Vocabulary Learning Hub. For the wider role of grammar, context, sentence force and communication in English, the protected How English Works core provides the appropriate foundation. Those owners remain independent; this article focuses only on the specialist TBX interchange, validation and terminology-governance workflow.

Final practical principle: export success is not semantic success. Keep concept identity, document the dialect, preserve the status of each designation, test the receiver and prove that a later writer or reader will not be guided toward a different meaning because the terminology moved. A usable terminological exchange is one in which people can still answer what a concept means, why a term was approved, where the evidence came from and what to do when the next ambiguity appears.

Architecture reference: EDKSG-TRANS-MASTER-WORLD-1040. Educational and technical interpretation; fictional diagnostic examples; no legal, clinical or software-certification claim.