Skip to content

Latest commit

 

History

History
37 lines (22 loc) · 1.86 KB

README.md

File metadata and controls

37 lines (22 loc) · 1.86 KB

COAR Vocabularies

The COAR Vocabularies are a standardised way to describe some aspects of content held in Open Access repositories. They are organised using SKOS (Simple Knowledge Organisation System).

The vocabulatries are being used in a variety of metadata profiles such as RIOXX3.

In this Git repository are a set of resources to create relevent EPrints files for the vocabularies, using the N-triples representation of each vocabulary e.g. https://vocabularies.coar-repositories.org/version_types/1.1/version_types.nt

The script will output:

  • a namedset file for each vocabulary
  • language specific phrase files for each vocabulary, in each language represented in the N-triple file.
  • [maybe at some point...] a map of URIs/phrases to be included in an XSLT for the OAI-PMH interface.

Processing the SKOS

NOTE: CURRENTLY THIS IS NON-STANDARD! Running a bin script from epm directory!!!

Each N-Triples SKOS file contains multiple language variations. Not all of these are relevant to a given repository.

I'm still working out the best way to deploy these to a repository. The namedsets files are common to all languages.

One option is to cache the lang files in a 'sources' directory and have an EPM script copy relevant ones into the appropriate location.

Currently running:

bin/process_coar_vocabs.pl

will output files into files/output/cfg/namedsets/... and files/output/cfg/lang/[LANG]/phrases/....

References