Skip to content
Property · DCMI Metadata Terms: properties

dct:identifier: meaning, literal range and examples for records

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

An unambiguous reference to the resource within a given context.

Recommended practice is to identify the resource by means of a string conforming to an identification system. Examples include International Standard Book Number (ISBN), Digital Object Identifier (DOI), and Uniform Resource Name (URN). Persistent identifiers should be provided as HTTP URIs.

In DCAT records and other Turtle files the property is usually written dct:identifier; in library documentation, dcterms:identifier. Both are http://purl.org/dc/terms/identifier. It carries the same definition as the original dc:identifier, an unambiguous reference to the resource within a given context, and it is formally a sub-property of it. The difference that matters to developers is a single line in the specification: the range is rdfs:Literal.

Why the value is a string

Linked-data newcomers often expect an identifier property to point at a URI. dct:identifier does not. Its values are strings: an ISBN, a DOI written as a resolver URL, a catalog number, an agency reference code.

The reasoning becomes clear once the triple is drawn out. In RDF the resource is already named by the subject of every statement about it. <https://example.org/id/dataset/rainfall-2023> is the identifier in the RDF sense. What dct:identifier adds are the other names the resource goes by in other systems, written the way those systems write them. A DOI given as "https://doi.org/10.5555/rain.2023" is a string the resource is known by, not a link to another node.

Two consequences follow. A URI in quotes is correct here even though it looks like it should be a link. And using dct:identifier with an rdf:resource or angle-bracket URI value produces data that conflicts with the declared range, which strict validators flag.

The subject URI question

A recurring debate is whether dct:identifier should repeat the subject URI. DCAT profiles differ. Some European application profiles recommend it, so that a portal reading only literal values still sees a stable identifier; others leave it optional. Repeating it costs one triple and harms nothing. The harm comes from repeating a different URI by accident, for instance a landing page address that does not match the subject, which leaves consumers with two candidates and no rule for choosing.

The opposite mistake also turns up: a record whose subject is a blank node or a temporary URI minted by an export script, with the only stable identifier hidden in a dct:identifier literal. Harvesters then create a new node on every run. Minting a stable subject URI from the persistent identifier avoids that duplication at the source.

Recording the scheme

Simple string values do not say which scheme they belong to. “HYD-RF-2023” means something to a hydrology agency and nothing elsewhere. Three common approaches:

Approach Example
Self-describing form https://doi.org/10.5555/rain.2023
URN with scheme prefix urn:isbn:9780000000002
Structured node adms:identifier pointing to a node with notation and scheme

The third option comes from the ADMS vocabulary, which is used alongside DCAT in European data portals. It lets a record say “this string is an identifier issued by this agency under this scheme” without bending the DCMI property.

Identifiers and relationships

An identifier names the resource described by the record, not a related one. When a dataset exists in CSV and NetCDF versions, each version has its own identifier, and the link between them goes through dct:isFormatOf or dct:relation. When a digital file was made from a printed original, the original’s identifier belongs in dct:source.

Mixing these up is easy in workflows that copy metadata from a parent record. A scanned book that inherits the print ISBN as its own identifier will be merged with the print edition by every system that deduplicates on ISBN, which is exactly the behavior identifiers are meant to support, applied to the wrong pair of resources. Ensuring each manifestation carries only its own numbers keeps that matching useful, and it is also what makes dct:title and identifier together a reliable basis for citation.

Example in RDF

@prefix dct: <http://purl.org/dc/terms/> .

<https://example.org/id/dataset/rainfall-2023>
    dct:title "Daily rainfall, 2023"@en ;
    dct:identifier "https://doi.org/10.5555/rain.2023" ;
    dct:identifier "HYD-RF-2023" ;
    dct:isFormatOf <https://example.org/id/dataset/rainfall-2023-netcdf> .

Questions about dcterms:identifier

What is dct:identifier?

dct:identifier is the prefixed form of dcterms:identifier, the DCMI Metadata Terms property for an unambiguous reference to the resource within a given context. It is a sub-property of the original dc:identifier and shares its definition.

Why is the range of dcterms:identifier a literal?

The property records identifier strings as they are written in their scheme, such as an ISBN or a DOI. The resource's own URI is expressed as the subject of the RDF statements, so the identifier property does not need to point at a resource.

Should dct:identifier repeat the subject URI?

It may, and some profiles ask for it so that the identifier is visible to tools that only read literal values. It is not required, and repeating it adds nothing when the URI already identifies the resource unambiguously.

How is the identifier scheme recorded?

dcterms:identifier itself has no attribute for the scheme. Writing identifiers in self-describing forms such as resolver URLs or URNs is one answer. Another is adms:identifier, which points to a node carrying the notation and the issuing scheme.

Can dct:identifier hold several values?

Yes. A dataset can have a DOI, an agency reference number and a project code at once. Each goes in its own statement, and the persistent identifier is usually listed first in serializations meant for people to read.

More from DCMI Metadata Terms: properties

All 13 in this group