To translate ISO 639:2023 language codes, language names and localization metadata correctly, first separate a stable machine identifier from a visible name that a human reader needs. An ISO language code is not a country, a script, a website locale or a translation of cultural identity. This guide explains how to preserve the key while choosing accurate, respectful and usable multilingual labels.
Readers searching for ISO 639 language code translation, endonym and exonym differences, language names in multilingual catalogues, BCP 47 tags, locale metadata and website language selectors often have a practical failure to repair. A translated label may be fluent while the language switcher routes to the wrong page, the library record loses its historical aliases, or the CMS no longer recognizes its original key. These are data-model problems as well as language problems.
The quick solution is to identify the code authority, freeze protected identifiers, translate only the correct display fields, verify script and regional conventions under the appropriate separate specification, and test the rendered interface with readers. This focused article walks through the logic, a set of worked decisions, bilingual review, FAQs and a four-week learning programme. It connects to the site’s protected master translation and vocabulary resources without replacing them.
Quick answer: protect the code, clarify the label
- Identify the authoritative code list and its current edition.
- Freeze machine keys that serve as identifiers.
- Translate reader-facing labels only under an explicit naming convention.
- Differentiate language, script, territory and locale.
- Validate BCP 47 web tags using the appropriate separate registry.
- Test actual content, routing and accessibility before release.
What the published ISO document covers
The official ISO 639:2023 catalogue describes a code for individual languages and language groups, rules for language identifiers and reference names, and the system maintained by the ISO 639 Maintenance Agency. It does not serve as a complete rule book for every web locale or country’s formatting customs. The edition was published in November 2023. Consult the primary record when a naming or coding decision must be current.
Primary references: ISO 639:2023 official catalogue; W3C explanation of web language tags. This independent educational guide does not reproduce licensed standard text.
A real-world starting problem
Imagine a publisher importing textbooks from several countries. One file describes language in English, another uses community names, and another places an apparent country code in a field labeled language. The temptation is to tidy the data by translating every field. But this approach destroys useful distinctions. The responsible editor checks the schema and registry before editing, then constructs a display layer that helps people without compromising record identity.
Chapter 1. A language is not a nationality
A person can use the same language across multiple countries and may use several languages in one country. A country flag is therefore a weak substitute for a language label. Verify the content and user choice rather than deducing language from residence, citizenship or market. If a page serves several regional forms, state that editorial convention explicitly.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 2. Identifiers are different from display names
The source schema may contain an uneditable code and a separate human-readable field. An identifier is meant to join records or select content; the label is meant to help people understand an option. Translating both as ordinary prose can break a registry join. Trace the data contract and write a rule for each field before editing.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 3. Endonyms and exonyms
An endonym is a community’s own name for a language; an exonym is a name used in another language. A language chooser may benefit from showing native names, while an English catalogue may use English descriptive names. Consider dual labels or search aliases when readers use both. Never infer that a preferred local term is a universal replacement across all contexts.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 4. Script and language must not be conflated
A writing system is not a geographic region or a separate language identity. BCP 47 language tags can include script subtags, but ISO 639 alone is not the entire BCP 47 system. Chinese content written in Simplified or Traditional characters illustrates the risk of translating script distinctions as country names. Use the appropriate specification and actual audience requirements.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 5. Region is a targeted qualifier
A region may be relevant to an interface, spelling convention or recorded voice, but an unqualified language should not automatically receive a country code. W3C guidance recommends using the shortest suitable language tag. Test whether the distinction changes the service delivered, not whether a longer code looks more professional.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 6. Reference names and search aliases
A library often requires a controlled record identity alongside historical spellings, local terms and alternative cataloguing labels. Users may search using names absent from the primary display field. Search aliases are helpful if attached to the same controlled record; inventing new identifiers for every synonym is not. Document the preferred name and its evidence.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 7. The meaning of a legacy code
Old systems may use a retired code or local shorthand that must be mapped to newer data structures. A migration is a technical task separate from a simple translation. Preserve the original value, the mapped value, effective date, registry and any exceptional uses. Do not discard an archival quotation simply because contemporary metadata now displays another name.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 8. Language tags on web pages
A multilingual page may contain text in more than one language. Where supported, identify language at the relevant HTML or XML span instead of assuming the whole document uses one speech variety. Assistive technologies can use these declarations for pronunciation. The visible translation and the invisible language metadata should agree as far as their formats permit.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 9. The role of accessibility
Language choosers must be discoverable and understandable to keyboard, mobile and screen-reader users. Icons alone can hide real linguistic differences, and an English-only alphabetical order can bury non-Latin names. Show clear labels, provide searchable alternatives and test focus behavior. A beautiful translation that cannot be selected or pronounced properly has not solved the user problem.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Chapter 10. Release and audit
Keep a source-to-target record of edited fields, unedited identifiers, terminology decisions and exceptions. Validate tags against their proper registry and test routing, fallback and sorting. Then ask a bilingual reader whether the labels actually identify the content that loads. Mechanical checks and human comprehension checks are complementary, not interchangeable.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
In this particular workflow, document whether the decision changes an identifier, a label, a reader’s interpretation or only presentation. Put a source reference beside the decision, then test a counterexample: could two different languages share a national context, could one language use two writing systems, or could a historical name still be a valid search alias? If the answer is yes, a single unqualified replacement is usually inadequate.
Thirty worked translation decisions
These scenarios are invented teaching cases rather than official example records from the standard. They focus on the editorial decision, the plausible failure, the defensible alternative and the check that makes the result more reliable.
Case 1: A CMS stores ‘fr’ as a page language key
The source situation is this: A CMS stores ‘fr’ as a page language key. A hurried editor might replace ‘fr’ with the word French. A safer choice is to retain the key and translate only the display name. The application expects a token for linking documents, while people need a meaningful name in the menu. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 2: A German menu lists ‘Deutsch’
The source situation is this: A German menu lists ‘Deutsch’. A hurried editor might replace it universally with German. A safer choice is to keep Deutsch in an endonym-based selector. Native-name selection helps a speaker recognize the language without already reading the interface language. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 3: A catalogue uses the English label Spanish
The source situation is this: A catalogue uses the English label Spanish. A hurried editor might assume Español is a wrong spelling. A safer choice is to use the established endonym or exonym convention. A display label is not an error because it uses a different naming language. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 4: A tag is written ‘en-US’
The source situation is this: A tag is written ‘en-US’. A hurried editor might claim ISO 639 alone specifies the whole tag. A safer choice is to interpret the language and region subtags under BCP 47. A combined web tag needs the syntax and registry that define its additional fields. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 5: A file is marked ‘zh-Hans’
The source situation is this: A file is marked ‘zh-Hans’. A hurried editor might call it Chinese from mainland China. A safer choice is to preserve the Simplified-script distinction. Writing system and location are not synonyms, even if a local market often uses one script. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 6: Another file is marked ‘zh-Hant’
The source situation is this: Another file is marked ‘zh-Hant’. A hurried editor might rename it Taiwanese. A safer choice is to retain the Traditional-script meaning. A script can be used across communities and does not by itself name the spoken variety. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 7: A site uses ‘es-419’
The source situation is this: A site uses ‘es-419’. A hurried editor might assume the digits refer to software release 419. A safer choice is to read the regional subtag according to BCP 47. The number in a language tag can designate a region, not a product version. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 8: A request says ‘Arabic for everyone’
The source situation is this: A request says ‘Arabic for everyone’. A hurried editor might promise a single locale fits every reader. A safer choice is to ask about variety, terminology and audience. The language identifier alone does not determine editorial register, legal needs or language-use preferences. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 9: An export field contains ‘PL’
The source situation is this: An export field contains ‘PL’. A hurried editor might translate it automatically as Polish. A safer choice is to inspect the schema and source authority first. Those letters might denote a country, layer, category or other nonlinguistic code. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 10: A field is called language_name_en
The source situation is this: A field is called language_name_en. A hurried editor might replace the English name with a native-script form. A safer choice is to keep the field’s role and use a separate localized name. Schema names may promise a specific representation to consuming applications. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 11: A recording is labeled Mandarin
The source situation is this: A recording is labeled Mandarin. A hurried editor might replace it with an undifferentiated Chinese option. A safer choice is to retain a useful spoken-language distinction. An audio selection must reflect what a listener can understand. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 12: An archive has an older language name
The source situation is this: An archive has an older language name. A hurried editor might erase every historical form. A safer choice is to use controlled current labels and historical search aliases. Older bibliographies and quotations should remain searchable and traceable. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 13: A subtitle file inherits the player’s language
The source situation is this: A subtitle file inherits the player’s language. A hurried editor might assume the spoken track has the same language. A safer choice is to identify the audio or signed content independently. A player interface and a transcript are different content layers. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 14: An interview switches languages mid-sentence
The source situation is this: An interview switches languages mid-sentence. A hurried editor might tag every passage as one uniform language. A safer choice is to document multilingual spans where the platform supports them. Real speech can contain code-switching and mixed-language turns. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 15: A Portuguese-language download has no region
The source situation is this: A Portuguese-language download has no region. A hurried editor might add a region on instinct. A safer choice is to verify whether an audience distinction is actually needed. A generic label is not evidence for a particular national standard or spelling style. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 16: A Tamil teaching document uses English examples
The source situation is this: A Tamil teaching document uses English examples. A hurried editor might label all content monolingual Tamil. A safer choice is to reflect the actual bilingual instructional content. Language identity should follow the text, not only its heading. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 17: A site uses a national flag as the sole language selector
The source situation is this: A site uses a national flag as the sole language selector. A hurried editor might substitute the country name for the language. A safer choice is to provide the written language label. One language can cross borders and a flag is not a semantic language identifier. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 18: A list contains three-letter identifiers
The source situation is this: A list contains three-letter identifiers. A hurried editor might expand them using a casual blog table. A safer choice is to check the applicable authoritative registry and code set. Historic variants and local application contracts can make informal lists unsafe. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 19: A vendor invents a short code for a dialect
The source situation is this: A vendor invents a short code for a dialect. A hurried editor might describe it as an official ISO assignment. A safer choice is to label it as a project-specific alias unless verified. Locally useful conventions should not be passed off as registered standards. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 20: A course catalogue says ‘Sign Language’
The source situation is this: A course catalogue says ‘Sign Language’. A hurried editor might treat every sign language as one universal language. A safer choice is to identify the specific signed language offered. Signed languages are distinct and audience selection depends on which one is taught. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 21: An archive stores English and French descriptions
The source situation is this: An archive stores English and French descriptions. A hurried editor might merge their provenance into one anonymous statement. A safer choice is to retain source, creator and date for each. Translated records may differ in their original authorship and level of detail. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 22: A translation memory changes a historical name
The source situation is this: A translation memory changes a historical name. A hurried editor might rewrite quoted historical documents silently. A safer choice is to separate archival transcription from normalized metadata. The archive may need to preserve evidence of earlier usage while updating reader-facing search. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 23: A clinic form infers preferred language from birthplace
The source situation is this: A clinic form infers preferred language from birthplace. A hurried editor might accept that automatic inference. A safer choice is to ask for actual language preference. Country of birth does not determine language competence or communication needs. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 24: A portal has a fallback locale
The source situation is this: A portal has a fallback locale. A hurried editor might assume every missing translation is in that language. A safer choice is to describe actual fallback behavior and content language. Software preference and loaded content are not necessarily the same thing. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 25: An unfamiliar abbreviation appears
The source situation is this: An unfamiliar abbreviation appears. A hurried editor might delete the material as invalid. A safer choice is to research the originating taxonomy and document mapping. A nonstandard entry may need reconciliation rather than erasure. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 26: A page includes a French quotation
The source situation is this: A page includes a French quotation. A hurried editor might label every span as English. A safer choice is to mark language changes where the document format supports them. Text-span metadata helps pronunciation and indexing when used correctly. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Start by identifying the role of this field in the actual source system. Ask which part the reader sees, which part the software reads, and who is authorized to change either. A translation that preserves this separation can be checked independently of stylistic choices. Record the original token, its category and the intended audience in the review log.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 27: A menu label changes between releases
The source situation is this: A menu label changes between releases. A hurried editor might check only spelling in the new screenshot. A safer choice is to test routing, sorting and screen-reader behavior. The visible label may participate in an interactive language-selection process. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Next test how the sentence travels into a second format. A spreadsheet column can include context lost in a dropdown, catalogue or caption. If the critical distinction disappears in the smaller interface, shorten surrounding prose rather than removing the distinction. The receiving application and the human reader both deserve a clear representation of the same source fact.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 28: A glossary translates locale as language
The source situation is this: A glossary translates locale as language. A hurried editor might treat those concepts as identical. A safer choice is to define the broader locale configuration used by the application. Number formatting, collation and regional settings may go beyond the language code. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Make the rejected version explicit during review. It helps colleagues see not merely which wording is preferred but what factual or structural mistake it would introduce. Save this reason in a controlled terminology record. A future translator then has an explanation rather than an unexplained list of approved words.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 29: A dataset stores several synonyms
The source situation is this: A dataset stores several synonyms. A hurried editor might turn every synonym into a separate identity. A safer choice is to keep aliases attached to the controlled language identity. Search coverage can improve without proliferating conflicting keys. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
When there is genuine ambiguity, do not conceal it behind fluent prose. Mark the alternatives, consult the registered authority or source author and explain any outstanding choice to the recipient. A documented uncertainty is more reliable than a confident but invented linguistic identity.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
Case 30: A project states ‘all languages validated’
The source situation is this: A project states ‘all languages validated’. A hurried editor might treat schema checks as proof of linguistic accuracy. A safer choice is to separate code validation from content review. Technical conformance does not prove that a person can understand the translated page. The correction is not merely a prettier wording. It protects a relationship between information and the people or systems interpreting it.
Finally ask a reader to use the translated artifact for its real purpose: find the language, select the track, open the page or retrieve the record. If the correct destination cannot be reached or recognized, improve the metadata and labels together. The user experience is the final test of a successful translation.
For verification, compare the target record with the original source and the authority appropriate to that field. Check whether the visible label and stored code still point to the correct item. Where an endonym, script or regional convention is disputed, document the intended audience and review with a qualified speaker. Keep the source value available for audit instead of silently overwriting it.
A repeatable editorial release workflow
Step 1
Receive the data dictionary, target languages, content format and expected users. Identify the authoritative code authority and web-tag requirements before translating a sample.
Step 2
Create a field register marking immutable identifiers, human labels, search aliases, script fields, region fields, effective dates and unresolved mappings. Do not treat all strings as ordinary prose.
Step 3
Build an evidence-based termbase with a preferred label, definition, audience, source authority, historical aliases and a sentence illustrating correct use.
Step 4
Translate labels, headings and explanatory text while leaving controlled identifiers unchanged. Where a field combines machine identity and visible text, separate it rather than modifying the key.
Step 5
Validate output tags and keys mechanically, then test language switching, sorting, fallback and accessibility in the actual recipient environment.
Step 6
Have speakers review reader-facing choices, particularly community names, signed-language distinctions, regional variants and labels whose English equivalent may carry unwanted assumptions.
Step 7
Release a versioned manifest of sources and changes. Correct failures through a controlled mapping update and keep prior values where archival traceability is required.
Frequently asked questions
Does ISO 639 define all locale settings?
No. Its focus is language identification and reference names. Application locale behavior and web tagging draw on other systems, so inspect the actual consumer specification.
Should I replace language codes with English names?
Not when those codes are machine identifiers. Keep the key and use an appropriate human-readable label field.
Is a language the same as a country?
No. Languages cross borders and countries contain multiple languages. A country flag is not a reliable universal language selector.
What is the difference between an endonym and an exonym?
An endonym is a community’s own name; an exonym is a name used from another language. Both can serve useful purposes depending on the audience.
Can I put a script in an ISO 639 code?
Do not confuse the language code with an entire web tag. Consult the separate language-tag specification when script or regional subtags are necessary.
What if a legacy catalogue contains an old name?
Preserve traceability, record the current controlled representation and maintain search aliases rather than erasing archival evidence.
Will machine translation solve inconsistent metadata?
No. It can propose descriptive text, but code validation, registry mapping and user testing require separate checks.
How should I report a disputed label?
Record the source, proposed choices, affected audience, reviewer and unresolved question. Avoid publishing a controversial change as an unquestioned technical correction.
Four-week practice plan
Week 1
Week 1: Collect twenty metadata fields from a multilingual learning resource. Classify keys, labels, scripts, territories and languages. Verify ten assignments against their documented authorities.
Week 2
Week 2: Build a controlled glossary of endonyms, exonyms, scripts, ambiguous abbreviations and aliases. Explain why a code must not be translated as ordinary prose.
Week 3
Week 3: Create a miniature language selector, with native-name labels and accessible search. Test a real page route, a fallback route, spelling and correct spoken-language tags.
Week 4
Week 4: Invite independent readers to find content using both native and English terms. Correct any failed mappings, log the root cause and preserve the source record.
Connections to the protected learning ecosystem
This specialist page supports, rather than competes with, Master Art of Translation (EDKSG-TRANS-MASTER-WORLD-000), the Vocabulary Learning Hub and How English Works — Translation and Meaning Across Languages. The master covers general translation architecture; this article addresses a specific code-and-label handover.
The final lesson is that the human-readable name can be revised for clarity while the controlled identity remains traceable. A responsible translation protects both sides of that relationship. The examples are explanatory, not a substitute for checking the licensed standard or current maintenance records in a real implementation.