Dublin Core

Dublin Core is the simplest and most widespread vocabulary for describing resources: fifteen generic elements — title, creator, date, and so on — that almost any system can produce and understand. It takes its name from Dublin, Ohio, where it was defined at a workshop hosted by OCLC in 1995, and is maintained by the Dublin Core Metadata Initiative (DCMI). The element set is also an international standard: ISO 15836, ANSI/NISO Z39.85 and IETF RFC 5013.

The fifteen elements

elementmeaning
titlea name given to the resource
creatoran entity primarily responsible for making it — the authors
contributoran entity that contributed to it — editors, translators, …
publisheran entity that made it available
datea point or period of time in its life cycle
typeits nature or genre
formatits file format, medium or dimensions
identifieran unambiguous reference to it — URL, DOI, ISBN
sourcea resource it is derived from — for an article, often the citation of its venue
relationa related resource
languagethe language of its content
subjectits topic — keywords, classification codes
descriptionan account of it — typically the abstract
coverageits spatial or temporal scope
rightsinformation about rights held in and over it — licence, access

Every element is optional and repeatable, and its value is a plain string. This is Simple Dublin Core. DCMI Metadata Terms (the dcterms: namespace) refine it into Qualified Dublin Core, with more specific properties such as abstract, issued, isPartOf or license, and with controlled vocabularies such as the DCMI Type Vocabulary (Text, Dataset, Image, …).

Where it is used

  • OAI-PMH: every repository must offer its records in Dublin Core, as oai_dc. It is the common denominator of metadata harvesting.
  • Web pages: <meta name="DC.title" …> tags in the page head, read by some repositories and library harvesters.
  • Inside richer formats: OpenAIRE's oai_openaire, EPUB package metadata and many application profiles reuse Dublin Core elements next to their own.
  • Linked data: Dublin Core terms are among the most used properties in RDF.

Limits, and the info:eu-repo convention

Simple Dublin Core is flat. A creator is a string, with no given and family name, no ORCID iD, no affiliation; date does not say which date; identifier does not say which kind; source is free text. Harvesters must guess. Hence two remedies:

  • Conventions on values. European repositories, following the DRIVER and OpenAIRE guidelines, put controlled URIs into the plain elements: info:eu-repo/semantics/article in type, info:eu-repo/semantics/openAccess in rights, info:eu-repo/grantAgreement/EC/H2020/… in relation.
  • Richer formats, offered next to oai_dc: DataCite, MODS, MARC, and for scholarly literature OpenAIRE's oai_openaire.

See also