Skip to content
Class · DCMI classes and encoding schemes

DCT LinguisticSystem: the class behind dcterms:language, with examples

Published Formatory editorial team
http://purl.org/dc/terms/LinguisticSystem

A system of signs, symbols, sounds, gestures, or rules used in communication.

Examples include written, spoken, sign, and computer languages.

The class name sounds academic, and that is deliberate. dcterms:LinguisticSystem was chosen so that the range of dcterms:language would not quietly mean “a natural language spoken by people”. Its definition is “a system of signs, symbols, sounds, gestures, or rules used in communication”, and the official comment lists the examples: written, spoken, sign and computer languages.

Why language needed a class

When DCMI moved from the fifteen simple elements to the richer DCMI Metadata Terms, each property was given a formal range so that linked data validators could check values. For dcterms:language the obvious candidate, “Language”, carried too much everyday meaning. A name like LinguisticSystem keeps the door open to British Sign Language, to Klingon, to Esperanto, and to COBOL, all of which a collection might need to record.

That breadth is useful in digital preservation. A source code archive cares deeply about which programming language a file is written in, because it determines which compiler or interpreter can still run it. Format registries record this for format specifications too; the C++ entry is one such value.

How values are supplied

In practice almost nobody types a resource as dcterms:LinguisticSystem directly. Values arrive in one of three shapes:

Shape Example Notes
Typed code literal "pt-BR"^^dcterms:RFC5646 most common in current profiles
Code literal from ISO 639 "por"^^dcterms:ISO639-3 three-letter, no region
URI of a language resource <http://id.loc.gov/vocabulary/iso639-2/por> resolves to a description

The RDFS semantics of rdfs:range mean that whatever sits in the object position of dcterms:language is inferred to be a LinguisticSystem. Nobody has to assert it. A typed literal is a slightly awkward fit for that inference, which is one reason linked data projects lean toward language URIs. The Library of Congress publishes URIs for ISO 639-1 and ISO 639-2 codes at id.loc.gov, and Lexvo.org offers URIs for ISO 639-3 codes, so a language resource rarely has to be minted locally.

See dcterms:RFC5646 for tag syntax and dcterms:ISO639-3 for three-letter codes.

Edge cases worth deciding in advance

Sign languages are frequently left out of records because the cataloger reaches for a spoken-language default, often English, even when the recording contains no spoken English at all. ISO 639-3 has individual codes for many of them, for example ase for American Sign Language and bfi for British Sign Language, and RFC 5646 accepts those as primary subtags.

Notation systems are another grey zone. Music notation, chemical formulae and mathematical markup are communication systems in a loose sense, but most profiles do not record them under dcterms:language, and doing so tends to confuse users who filter by language. A note in the local cataloging manual saves repeated debate.

Constructed languages are the opposite case: they are plainly languages, and ISO 639 covers several. Esperanto has eo and epo, Klingon tlh, Lojban jbo. A fan fiction archive or a linguistics corpus can record them in exactly the same way as English or Swahili, and they behave normally in any language facet.

For resources without any linguistic content, a photograph with no caption, an instrumental recording, the code zxx exists. It is different from an empty field, which only says the language was never recorded.

Relation to the simple element

The legacy dc:language element has no formal range, so it accepts anything. When simple records are upgraded to DCMI terms, a quick audit of the distinct values in that field usually turns up language names in several languages (“German”, “Deutsch”, “allemand”) that need to be mapped to a single code before the class constraint means anything.

Example in RDF

@prefix dcterms: <http://purl.org/dc/terms/> .
@prefix rdfs:    <http://www.w3.org/2000/01/rdf-schema#> .

<https://example.org/video/sign-lecture-12.mp4>
    dcterms:language <http://id.loc.gov/vocabulary/iso639-2/sgn> ,
                     "en"^^dcterms:RFC5646 .

<https://example.org/code/parser-v2.zip>
    dcterms:language <https://example.org/langs/python3> .

<https://example.org/langs/python3> a dcterms:LinguisticSystem ;
    rdfs:label "Python 3" .

Questions about dcterms:LinguisticSystem

What is dcterms:LinguisticSystem?

It is the DCMI class for languages in the widest sense: any system of signs, symbols, sounds, gestures or rules used to communicate. It is declared as the range of dcterms:language, so every value of that property is understood to be a linguistic system.

Does LinguisticSystem include programming languages?

Yes. The official comment names computer languages alongside written, spoken and sign languages. A source code package can therefore carry dcterms:language pointing to a resource that stands for Python, C++ or Fortran.

Do I have to create a LinguisticSystem resource for every language?

No. A typed literal such as 'fr' marked as an RFC 5646 tag, or a code from ISO 639, is the usual way to supply the value. Separate resources make sense when no standard code exists, as with many programming or constructed languages.

Can a resource have more than one language?

Yes. dcterms:language is repeatable. A bilingual edition gets one statement per language. A video with signed narration and English captions can list the sign language and English as separate values.

How do I say a resource has no language?

Use the code zxx from ISO 639-2, which means no linguistic content. Leaving the field empty says something different: that nobody recorded the language, which is not the same as there being none.

More from DCMI classes and encoding schemes

All 10 in this group