(I'm not sure whether this is the best list to post this question to: feel free to suggest other places I should try!)

At Durham University we are migrating our digital repositories (research publications, theses, digitised special collections etc.) from a variety of platforms to Hydra/Fedora.  In Fedora 4 the natural way of storing descriptive metadata is RDF.  We are therefore trying to identify a good selection of vocabularies to cover our needs.  The Sufia application (built on Hydra) tends to use things like DC and FOAF by default, but we could use any combination of anything, or, indeed, locally-defined predicates.

Our first application of the new repository was research data, and we struggled to find suitable predicates for things like the Funder of the research, and the Affiliation of the creators of the data.  For these and a few others we temporarily made up our own predicates.  Then I became aware of the RDF vocabulary for RDA, and the parallel work going on with BibFrame's RDF representation (I won't ask why we need two standards!).  These looked quite promising, as we could identify predicates for several of our "problem fields".

Now, however, I am looking at the example of journal articles. In line with our Converis CRIS system, we expect to model these in Fedora with an object representing the journal, and an object representing the article.  The article object needs to record the metadata relating to the published version of the article.  Attached to the object will be at least one further object carrying the author-final-version of the article which we are permitted to hold in the repository, but potentially there may be several related objects.

I would like to focus on the object representing the published article.  In particular we need to record the volume and issue of the journal, the start and end pages, and the year.  It is necessary to have these separate pieces of information articulated in separate fields because we will need to support export of the citation for EndNote, BiBTeX and OpenURL.

Mike Taylor, of course, wrote an excellent article on this problem:
https://reprog.wordpress.com/2010/09/03/bibliographic-data-part-2-dublin-cores-dirty-little-secret/

One of my colleagues has therefore gone through several possible vocabularies to see what we can come up with.  It seems that RDA has rdau:P60057 "has preferred citation" but that does not separate the volume, issue, etc. into separate fields.  BibFrame does not cover it yet.  Dublin Core is inadequate, as Mike Taylor points out.  We tried looking at CERIF but could not find an RDF representation.  RIOXX mainly reuses DC and does not cover these fields.  OpenAIRE ditto.

About the only vocabulary we have found so far that covers these fields is the Bibliographic Ontology: http://bibliontology.com/ but I am not sure how widely that has been adopted.

I had hoped that RDA might cover our needs: after all, it's meant to be a comprehensive system for cataloguing everything.  But I cannot see how you are meant to handle this fairly basic use case.  We're not against mixing our predicates from different vocabularies, but it would be nice to use only a few.

I'd be grateful for any thoughts from those more experienced in the ways of RDF vocabularies!

Thanks,

-- 
Matthew Phillips
Head of Digital and Bibliographic Services,
Durham University Library, Stockton Road, Durham, DH1 3LY
+44 (0)191 334 2941