Skip to content
Property · DCMI Metadata Terms: properties

dct:type in practice: classes, DCMI types and the rdf:type difference

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

The nature or genre of the resource.

Recommended practice is to use a controlled vocabulary such as the DCMI Type Vocabulary [DCMI-TYPE]. To describe the file format, physical medium, or dimensions of the resource, use the property Format.

Every RDF vocabulary user eventually meets two properties that both seem to say what something is: rdf:type and dct:type. They answer different questions, and the difference explains most of what is unusual about the DCMI property. dct:type, written dcterms:type in library documentation, records the nature or genre of the resource. Its definition matches the original dc:type element, of which it is a sub-property, and its range is rdfs:Class.

A range of rdfs:Class

That range means the value of dct:type is itself a class. The DCMI Type Vocabulary was built for this: each of its terms, from dcmitype:Collection to dcmitype:Text, is declared an rdfs:Class in the namespace http://purl.org/dc/dcmitype/. A conforming statement is therefore dct:type dcmitype:Dataset, a URI, and the bare string “Dataset” is a simple-Dublin-Core habit that does not match the range.

The design choice is subtle. Using a class as a value treats the class as an individual, a pattern sometimes called punning in OWL. It lets a record say “this resource is of the genre Dataset” without committing it to every inference that membership of the class would bring.

For anyone exporting to simple Dublin Core the URI form costs nothing. A crosswalk can reduce dcmitype:Dataset to the literal Dataset by taking the last path segment, which is exactly the string oai_dc consumers expect. Going the other way, from inconsistent literals to URIs, is the harder direction, so storing the URI internally is the safer default.

dct:type versus rdf:type

rdf:type dct:type
Purpose class membership in the data model descriptive genre
Effect on reasoning triggers inference none beyond range
Typical value dcat:Dataset, foaf:Person dcmitype:Dataset, a genre term

In practice both appear in the same record and often with different values. A DCAT catalog uses rdf:type dcat:Dataset because the DCAT model needs it, and it can add dct:type with a genre from a controlled list, for instance one distinguishing statistical tables from geospatial layers. For a format registry, rdf:type might place an entry in a modeling class while dct:type says the format is primarily an image or an audio container.

Using dct:type where the data model needs rdf:type is a frequent error in hand-written Turtle. Validators checking a DCAT profile will report the dataset as untyped, because they look for rdf:type (written a in Turtle) and ignore the DCMI property.

Beyond the twelve DCMI types

The DCMI list is short by design, and many collections need more detail. Sources commonly paired with it:

  • the COAR Resource Types vocabulary for repository outputs such as journal articles, theses and research data
  • the Library of Congress Genre/Form Terms for library and archival material
  • local SKOS schemes for domain-specific genres

The usual approach is to keep one DCMI type for broad interoperability and add a more specific genre as a second dct:type value, both as URIs. Search interfaces can then offer a coarse facet across all collections and a fine facet within each.

Type, format and subject

The usage comment draws the line with format explicitly: file format, physical medium and dimensions go in the Format property, so dct:format holds image/tiff, while dct:type holds StillImage. A media type never answers the type question, since a PDF can be a book, a map or a spreadsheet.

The other line runs to dct:subject. A collection of photographs of trains is of type StillImage with subject railways; a book about photography is of type Text with subject photography. Records that put “photographs” in the subject field and leave the type empty are common in older catalogs, and moving those values across is one of the cheaper cleanup tasks with a visible payoff in faceted search.

Example in RDF

@prefix dct: <http://purl.org/dc/terms/> .
@prefix dcmitype: <http://purl.org/dc/dcmitype/> .
@prefix dcat: <http://www.w3.org/ns/dcat#> .

<https://example.org/data/ferry-timetables>
    a dcat:Dataset ;
    dct:type dcmitype:Dataset ;
    dct:title "Ferry timetables, 2025 season"@en ;
    dct:format <https://www.iana.org/assignments/media-types/text/csv> .

Questions about dcterms:type

What is dct:type?

dct:type is the prefixed form of dcterms:type, the DCMI Metadata Terms property for the nature or genre of a resource. It is a sub-property of dc:type and is usually given a value from the DCMI Type Vocabulary, such as Text, StillImage or Dataset.

What is the range of dcterms:type?

The range is rdfs:Class. The value is therefore expected to be a class, identified by a URI, such as http://purl.org/dc/dcmitype/Dataset. A plain literal like Dataset does not fit the declared range.

What is the difference between dct:type and rdf:type?

rdf:type states class membership in the RDF data model and drives inference. dct:type records the nature or genre as descriptive metadata. Many records use both, often with different values: rdf:type for the modeling class, dct:type for a genre.

Can dct:type use terms outside the DCMI Type Vocabulary?

Yes. The DCMI list is a recommendation. Genre vocabularies, the COAR resource types or a local SKOS scheme can supply more specific values, ideally as URIs, with one DCMI type added where cross-collection consistency matters.

Is a file format a type?

No. The usage comment points format, medium and dimensions to the Format property. A TIFF file is described by dct:format; its dct:type depends on content, for example StillImage for a photograph or Text for a scanned letter.

More from DCMI Metadata Terms: properties

All 13 in this group