How do you translate art and architecture terminology accurately when a familiar word can describe several different materials, objects, styles or techniques? The Getty Art & Architecture Thesaurus, usually called AAT, provides structured concepts, terms, relationships and supporting information for describing visual art, architecture and cultural heritage. Multilingual cataloguing works best when different language labels point to the same carefully identified concept. It fails when an easy synonym hides a narrower technical meaning or makes two unrelated concepts appear identical.
People searching for Getty AAT multilingual labels, Art & Architecture Thesaurus translation, museum controlled vocabulary localization, AAT concept URIs, architecture term mapping or linked open data reconciliation need to understand the distinction between a term and a concept. A word is a linguistic expression used by a community; an AAT concept is an identified unit of meaning that may have preferred labels, alternate names, hierarchical neighbors, sources and scope notes. The central task is not to replace English words in a database. It is to preserve those conceptual boundaries while providing natural, culturally responsible labels in other languages.
This guide explains the Getty AAT’s role in multilingual museum and visual-resource records, from selecting concepts and handling homographs to querying hierarchies, mapping materials and maintaining linked data. It is written for curators, terminologists, translators, art historians, information professionals, developers and global heritage portals. Current Getty documentation also matters: in 2026 the organization points implementers toward Linked Open Data formats and records that its older XML web services and regularly refreshed XML and relational releases have been discontinued. Readers will learn how to make language choices while keeping a stable basis for searching, attribution and future correction.
Why a multilingual thesaurus is more than a translated glossary
Imagine a global museum portal that holds records from a dozen countries. Visitors use many everyday words for the same kind of material, and a single familiar word sometimes means several different things. One catalogue may call a process by its historic name; another may use an industry term. The Getty AAT gives the portal a way to anchor those labels to a shared concept, with documented context and relationships. The public interface can then be welcoming without replacing the carefully identified category used for research.
This separation has practical consequences. The page title can be short; the concept record must be precise. The search index may accept synonyms; the classification record must point to the intended identity. The translation team can disagree about a preferred expression; the source’s evidence and scope note remain available to resolve the question. The result should be a cataloguing system that becomes more accessible as languages are added rather than less trustworthy as terminology multiplies.
Thirty-two practical Getty AAT terminology and localization decisions
1. The concept is not the displayed word
A catalogue contains a common term such as print, but that word can refer to a reproduction or a particular artistic process. An AAT concept identifier defines a unit of meaning beyond one human language label. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
Translating a short English word with the first dictionary result can move the record into the wrong specialist category. Start from the concept record and its documented scope, then select a target-language term. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Search the selected concept ID and verify that the target label describes the same category. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
2. Persistent AAT concept URIs
A museum catalogue stores an AAT concept URI linked to a work type or technique. The URI establishes stable reference across languages and should not be rebuilt from a translated phrase. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
Replacing the concept ID with a new slug creates broken links and can split a single classification across systems. Keep the AAT identifier intact and attach approved language labels separately. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Reopen the Getty concept record and confirm that the object still points to the same identifier. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
3. Preferred and alternate terminology
An art historian’s preferred technical word may coexist with older spelling, regional usage or a public-friendly synonym. Synonyms help discovery but do not necessarily carry equal authority or suitability in every context. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
A translation system that picks any synonym without regard to status can generate inconsistent museum labels. Use the term status and usage context supplied by the vocabulary rather than relying only on frequency. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Check cataloguing forms and public search aliases for consistent selection. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
4. Homographs with different meanings
An everyday word can refer to a craft tool, an architectural part or a style depending on its surrounding context. Disambiguation requires concept scope and hierarchy, not simply matching the word form. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
Two concept records sharing a term can be accidentally merged by a spreadsheet deduplication rule. Reconcile against concept identifiers and review neighboring terms before assigning multilingual equivalents. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Inspect candidate matches whose strings are identical but whose conceptual parents differ. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
5. Scope notes guide translation
A Getty vocabulary concept can have explanatory scope notes that delimit its use. A good translation should reflect the defined scope even when the term itself is brief. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
A locally familiar term may be misleading if it includes objects or processes excluded by the concept. Read the entire concept note and compare real objects at the boundaries of its meaning. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Ask whether the target term would classify both a valid and an invalid example correctly. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
6. Hierarchies explain relationships
A material concept may have broader and narrower relatives in one or more structured contexts. Broader and narrower relations help users navigate between general and specialist descriptions. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
A target label can be too general, turning a particular material into its whole family. Choose the narrowest justified concept for cataloguing and provide a broader explanatory phrase only for accessible display. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Check whether browsing from the translated parent leads to the same child concepts. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
7. Polyhierarchy and multiple contexts
A specialized technique may be meaningfully connected to more than one organizing context in a thesaurus. A concept’s multiple relationships need not imply that it has several conflicting identities. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
Flattening a concept to one invented parent can hide relevant avenues of discovery. Keep the complete authorized hierarchy and select suitable display paths without changing classification. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Test several browse routes that should lead to one underlying concept. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
8. Facets are not interchangeable
AAT arranges terminology through broad semantic facets such as materials, activities, objects and styles. Facet context tells a cataloguer which kind of property is being recorded. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
An activity term entered as an object type produces records that remain searchable but are semantically confused. Map each metadata field to the correct kind of controlled concept before translation. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Review the concept facet and the field purpose for representative catalogue entries. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
9. Objects and work types
A museum can catalogue a textile, ceremonial vessel or piece of furniture using specialized object categories. These categories may be more precise than the general nouns visitors know. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
Translating a narrow object type as artifact removes distinctions vital to research and collection management. Retain the exact concept for indexing and add a plain-language explanation in public text. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Check that object-type filters return the same works after language changes. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
10. Materials are not techniques
Gold leaf, oil paint, glass and clay are materials, while gilding and firing describe actions or techniques. The same work may need separate controlled references for what it contains and how it was made. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
A fluent phrase that treats gilding as a material can corrupt technical queries. Map materials and activities to their proper fields and select terminology accordingly. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Search separately for gold leaf and gilding to test whether the intended distinctions survive. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
11. Styles and periods need context
A style label can describe historical, regional or intellectual patterns that lack exact one-word equivalents. A target-language label must preserve the source concept rather than imply a different movement. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
Translating a period name as a broad national era can erase important chronology and scholarly context. Use approved art-historical nomenclature, with explanatory notes where a direct equivalent is weak. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Check examples within the Getty concept’s scope rather than guessing from its English form. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
12. Roles and occupations
Terms such as restorer, engraver, architect and metalworker refer to different activities and professional roles. Role vocabulary clarifies what a named person did in relation to an object. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
A generalized word like artist may obscure participation in conservation, manufacture or documentation. Choose a role term at the documented level of detail and keep actor identities in their own authority records. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Review credits on objects with multiple participants and technical contributions. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
13. Cultural and historical designations
An older cataloguing term may reflect historical scholarship, community practice or outdated classification. Translating responsibly requires distinguishing documentary evidence from an endorsed contemporary public label. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
Replacing every older term can obscure the record’s history, while copying it without context can make it appear neutral. Preserve the source and provenance of historical language and seek expert or community guidance for public terminology. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Test whether the public page explains the difference between historical and current usage. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
14. Regional language variation
An object type may have different accepted names among speakers of the same broad language. Regional variants can support search without making the underlying concept change. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
Choosing only one local term as universal can make an international site sound alien or misleading. Record preferred regional forms where supported and allow alternate search terms under one concept ID. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Check actual cataloguing practice with relevant local institutions and audience samples. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
15. Singular, plural and grammatical form
A term may need inflection or changes in grammatical gender when embedded in a translated sentence. Language grammar influences display but should not become a new concept identifier. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
Automated assembly of isolated terms may produce unnatural or ambiguous phrases in context. Test vocabulary labels in complete sentences and consider separate display templates where languages require them. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Review singular and plural catalogue headings and field labels with fluent specialists. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
16. Script, diacritics and transliteration
An AAT term may contain scripts or diacritics that a local interface does not normally use. Unicode fidelity and correct script representation are part of reliable terminology. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
Removing diacritics or applying inconsistent transliteration can cause false duplicates or failed searches. Preserve the authoritative form and add searchable alternatives with clearly documented roles. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Test keyboard queries, normalization and screen-reader rendering where practical. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
17. Terminological evidence and sources
A vocabulary record can cite publications and contributors supporting a specialized concept or label. Those sources help a translator decide whether an unfamiliar term is documented or merely plausible. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
A translator’s clever neologism may appear authoritative if the source evidence is hidden. Consult the supporting references and document a provisional translation when an approved equivalent is absent. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Require review before a new specialist label becomes the preferred public term. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
18. Exact equivalence versus related meaning
Two language terms can overlap without having identical extensions in art history or material culture. Concept mapping should represent the strength of semantic correspondence honestly. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
Declaring an exact match when the target covers a broader tradition may wrongly merge many objects. Use qualified mappings or descriptive labels where exact equivalence cannot be defended. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Compare edge cases that belong to one concept but not the proposed target counterpart. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
19. Concept reconciliation in catalogues
A legacy museum database may use free-text material names with spelling variants and abbreviations. Reconciliation links those strings to controlled AAT concepts after verification. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
Bulk automatic matching by term alone can misidentify homographs and specialist techniques. Combine automated candidate suggestions with context, scope notes and human decisions. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Audit rejected matches and difficult examples before applying the mapping to the whole collection. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
20. Work type and iconography
A subject depicted in a painting is different from the physical kind of painting or its material. AAT work-type terminology should not automatically replace iconographic subject authorities. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
A painting of a ship is not itself a ship in the collection’s object-type facet. Keep depicted subject fields distinct from the classified object and consult appropriate subject vocabularies. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Compare search results for objects that are ships against paintings that depict ships. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
21. AAT versus ULAN
The Getty Union List of Artist Names supports authority work for individuals and organizations. AAT generic terms are not a substitute for stable identities of named artists. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
Entering an artist’s personal name as an AAT technique creates a malformed classification. Use the appropriate Getty vocabulary for the kind of entity and retain the relevant URI. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Review examples involving roles as generic concepts and named people as authority entities. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
22. AAT versus TGN
The Getty Thesaurus of Geographic Names relates to named geographical places. AAT concepts may include generic building or landscape categories without identifying a particular city. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
Using a generic urban-architecture term as a city identifier breaks geographical discovery. Map named locations to the appropriate geographic authority rather than confusing them with object types. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Test that location and work-type searches behave independently. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
23. AAT versus CONA and iconography resources
Getty vocabularies have different purposes for objects, people, places, generic concepts and iconographic subjects. A multilingual integration should choose the vocabulary appropriate to the question being asked. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
Trying to solve all forms of naming with AAT can lose the uniqueness of an artwork or person. Define the role of each authority and keep cross-links between them explicit. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Review a case combining a specific painting, its maker and a depicted subject. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
24. Linked Open Data instead of old XML feeds
Getty’s 2026 data-services guidance points implementers toward Linked Open Data and states that legacy XML web services are discontinued. A maintained translation workflow should use supported releases and identify the freshness of the downloaded data. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
An unattended XML poll may silently stop updating terms while a catalogue assumes it is current. Plan for RDF or JSON-linked records, documented refreshes and version checks. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Compare a sample of local concept data to the current authoritative online record. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
25. SPARQL and multilingual queries
A museum can query Getty linked data for concept information and labels used across languages. Language-aware retrieval should return suitable labels while preserving the concept URI. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
A query that filters only by English wording can miss a valid concept known through another label. Query stable URIs and implement language selection explicitly in the application. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Test same-concept results under multiple language preferences and document fallback behavior. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
26. JSON and RDF serialization
A linked-data dataset can express concepts in machine-readable formats that have structured keys and literal values. Translatable labels are distinct from namespaces, RDF predicates and parsing syntax. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
A global search-and-replace of terminology can accidentally change property names or invalidate serialization. Limit translation to designated language-bearing values and validate the output with appropriate parsers. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Inspect escaped characters, language tags and graph links in the target representation. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
27. Licensing and attribution
Getty states that its vocabulary data is available under the Open Data Commons Attribution License 1.0. Use rights and credit rules applicable to the data, not vague assumptions about free access. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
Copying a vocabulary table into a public product without proper attribution can undermine documented reuse terms. Link to the official license and cite Getty, contributors and relevant source records where required. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Review the public dataset documentation and credit page before release. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
28. Editorial changes and concept history
Vocabulary editors revise labels, merge or clarify records, and add terminology as scholarship develops. A concept’s current preferred term may differ from the form used in an earlier catalogue. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
An untracked update may cause one translation to contradict a stored historical annotation. Maintain release dates and a change-review process that separates vocabulary revision from prose translation. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Check important revised concepts and retest dependent catalogue facets. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
29. Missing target-language equivalents
Some specialist concepts have no established label in the requested language. A gap should be visible to editors rather than hidden behind a confident but invented equivalent. This is why the meaning of a controlled term must be checked against the vocabulary record rather than inferred from one display word.
An automatic guess can become widely copied simply because it appears in the preferred-term slot. Use the source term with a careful explanation or a clearly provisional label pending expert approval. Keep a note of the selected concept URI and the evidence that supports the target expression, especially when a curator must approve the terminology.
Record review status and test whether the portal misrepresents a fallback as an authoritative translation. A successful result is not merely fluent: a reader choosing another language should still be able to find the same kinds of objects and materials.
30. Search synonyms and false positives
Public visitors may search using broad everyday terms while curators need narrow specialist classifications. Search expansion can help readers find concepts without reclassifying the underlying objects. A multilingual thesaurus is useful precisely because labels can vary while the identified concept remains stable.
Aggressive synonym expansion can make a technique-specific query return unrelated artworks. Separate search aliases from authoritative indexing values and tune expansion by facet. Treat a proposed equivalent as a hypothesis to test, not as a dictionary fact. Compare the concept’s examples, scope and neighboring terms.
Inspect false positives and false negatives across languages using a curated test set. Test both cataloguing and public search. A label may look acceptable on the object page but create a misleading facet across a large collection.
31. Multilingual interface facets
A portal may show filters for material, technique, object type and period in several languages. The translated facet name and each concept label perform different informational jobs. An institution has to distinguish the term a visitor recognizes from the precise category a research system records.
A wrong facet heading can make otherwise correct concept values appear under an inaccurate category. Localize category labels and concept labels independently, preserving record-level controlled IDs. If simplification is helpful, add a short plain-language explanation instead of downgrading a specialist classification.
Check filter counts and category behavior after switching languages. Compare the actual linked identifiers before and after localization; a changed result set is a signal to investigate semantic drift.
32. Release governance and correction
A museum can ship a technically valid multilingual catalogue that nevertheless misclassifies an important object. Human oversight is necessary for culturally significant or technical terminology decisions. The same word can be a safe explanatory label in one sentence and an unsafe classification in another. Context determines the decision.
Fixing only the visible label without repairing a wrong concept link leaves search corruption in place. Track source data, mapping decisions, translation review and public corrections as separate stages. Record how the expression will be used, who reviewed it and whether it is preferred, alternate, historical or provisional.
Rerun critical queries and maintain a route for curators to challenge an incorrect mapping. Review examples that sit at the edge of the category as well as typical examples. The difficult objects reveal whether two languages really mean the same thing.
Worked problems: when a label damages a concept
A print is not every kind of printing
A catalogue uses a specialist printmaking term. An automatic target-language substitution turns it into a generic printing service category. Recover the documented concept, check its scope note, and use a technical equivalent or explanation rather than a misleading broad noun. Test relevant material and technique searches to ensure distinct works stay distinct. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
A medieval style becomes a modern national label
A historical style has regional scholarly use, but the target translation chooses a contemporary nation-name adjective. Preserve the period and cultural context as actually defined, and ask a specialist to review any contested terminology. Check that the selected concept still contains the example works used in the source catalogue. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
An artist name appears in the materials column
A malformed migration puts a person’s name where an AAT material term should be. Use the correct named-agent authority and preserve the material concept separately rather than inventing a material called by the artist. Inspect role and material facets after correction. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
Synonym matching merges two concepts
Two different concepts share a commonly used string, and the import process deduplicates by that string instead of a stable identifier. Restore the two concept URIs and their narrower scopes. Keep any ambiguous term only as context-dependent search assistance. Compare search results for both concepts and their hierarchical parents. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
A cultural term has no easy translation
A locally meaningful object or technique name lacks a standard equivalent in the target language. Keep the documented term with a concise explanation, and record that the solution is interpretive pending expert review. Do not promote an unsupported coinage into authoritative terminology simply to fill an empty field. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
A material becomes an activity
A translated field substitutes gilding where the structured metadata intends gold leaf. Distinguish the material present from the process of applying it, and correct the vocabulary link accordingly. Test a query for gold leaf across records produced by more than one technique. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
A 2025 XML integration stops refreshing
An institution relied on a Getty XML service discontinued in the infrastructure transition and continues to show stale preferred labels. Move to supported linked data workflows, keep update timestamps and independently compare selected concept records to Getty’s current service. Treat data freshness as an operational responsibility rather than pretending a translation is eternally finished. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
A free dataset is reused without credit
A public catalogue republishes terminology from a licensed dataset while omitting required source and contributor acknowledgment. Add the applicable Open Data Commons attribution and maintain traceable concept references. Review terms of reuse before redistribution and distinguish the vocabulary’s data license from any artwork image license. The example illustrates why reconciling concepts and translating terms are related but separate tasks, each requiring its own evidence and review.
A workflow for quality-assured multilingual terminology
Select a small sample of collection records with ambiguous object, style, material and technique names. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Identify the source AAT concept URI for each controlled term and preserve it as the identity anchor. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Check scope notes, alternate terminology, hierarchies, term status and cited sources before proposing target labels. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Separate an exact concept match from a broad explanation, an approximate equivalent or a search alias. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Keep artist identities, geographic names, specific works and iconographic subjects in their appropriate authority systems. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Record the language, regional usage, reviewer and status of each accepted or provisional term. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Validate linked-data serialization and preserve URI and predicate syntax rather than translating technical keys. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Use supported Getty data services and document data freshness; retired XML routes should not be assumed to update. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Test real catalogue facets and query counts in multiple languages using the same stable concept identifiers. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Publish attribution and a correction path, including review of culturally sensitive and disputed terms. Treat unusual cases as evidence that an automated rule needs revision. A small reviewed test set is cheaper than thousands of public misclassifications.
Frequently asked questions
What does the Getty AAT contain?
AAT provides structured generic concepts, terms and relationships for cataloguing and searching art, architecture and material culture.
Is an AAT term the same as its concept?
No. A concept is identified by a stable record; terms are language expressions that can represent it.
Can two concepts have similar words?
Yes. Homographs and overlapping terms require disambiguation through scope notes, contexts and identifiers.
Should translators change AAT concept IDs?
No. Keep the machine identifier unchanged and review the language label separately.
Why use scope notes?
They clarify the boundaries of specialist terms and help prevent attractive but inaccurate equivalents.
Does AAT identify individual artists?
AAT is for generic concepts; Getty’s ULAN is designed for named artists and related agents.
Does AAT identify particular cities?
Use the relevant geographic authority, such as Getty TGN, for named places rather than generic AAT place categories.
Are Getty’s old XML feeds still current?
Getty says its previous XML web services and regularly refreshed XML and relational release services are discontinued; consult its current linked-data guidance.
Can Getty Vocabulary data be reused?
Getty describes its vocabulary data as available under the Open Data Commons Attribution License 1.0, with attribution obligations.
What is the essential release test?
Check that a target-language label leads to the same concept and produces the same defensible classification as the source.