[OAI-implementers] rdf

Wed, 2 May 2001 19:29:17 -0500

Carl-

I agree that it's probably not worth going to a lot of effort solely to
have an RDF metadata format option on an OAI Provider site.  While the W3C
RDF specs contain numerous examples of how to encode DC metadata within RDF
structures, the RDF development community now seems to have a focus well
beyond simple metadata interchange.  A bit of a "religious war" between the
RDF Schema Language and XML Schema Language camps has developed.   OCLC's
involvement in RDF seems to have dropped off a little, there's still
ambiguity (and not enough good examples) about best practices for using RDF
in a metadata sharing application, it is harder to do, and it's not clear
to me what near-term role (if any) RDF will have with regard to metadata
for most of us, especially in regard to the type of bibliographic metadata
of prime interest to libraries and the like.  So I agree that these
ambiguities in combination with the difficulty level will likely inhibit
most OAI providers for now at least.  These factors also make it difficult
to develop anytime soon an OAI Harvesting application that could expect to
do a good job aggregating metadata provided in RDF -- which means even if
you offer RDF metadata format on your OAI Provider site, you may or may not
get any takers for your records in that metadata format.

That said, an OAI provider may have other reasons to store the metadata
that it reveals in RDF.  The metadata we're generating is multi-purposed,
i.e., it's designed to be used in a variety of applications both parochial
and collaborative.  We're experimenting with RDF and ancillary DC semantics
in order to experiment locally with different search architectures and
interfaces.  Since it's relatively easy to "dumb-down" complex XML metadata
records using XSLT (or other approaches), we've been able to generate the
OAI metadata records we reveal directly from our more complex RDF metadata
records with relatively little difficulty.  The RDF versions of our records
remain available for local system use and/or sharing by other
interoperability mechanisms, while just the DC content from the RDF records
goes into the OAI GetRecord and ListRecords responses.

Of course in this situation there's a natural temptation to want to reveal,
using OAI, our metadata in their most complex version (partly just to see
if it can be done).  We're at the stage where we're beginning to get a feel
for RDF and multiple namespaces (note, the example Tom references is still
very much a work in progress), and so naturally we've started thinking
about how it might be possible to reveal RDF records using OAI.  I think
it'll be a useful effort for us to try this, even though we don't really
expect it to go anywhere in the near term.

Tim Cole
Chair, Library Information Technology Committee
University of Illinois at Urbana-Champaign

----- Original Message -----
From: <lagoze@cs.cornell.edu>
To: <eric_morgan@ncsu.edu>; <oai-implementers@oaisrv.nsdl.cornell.edu>
Sent: Wednesday, May 02, 2001 2:56 PM
Subject: RE: [OAI-implementers] rdf

> Comments inserted below and some text deleted.
>
> > -----Original Message-----
> > From: Eric Lease Morgan [mailto:eric_morgan@ncsu.edu]
> > Sent: Wednesday, May 02, 2001 9:02 AM
> > To: oai-implementers@oaisrv.nsdl.cornell.edu
> > Subject: Re: [OAI-implementers] rdf
> >
> >
> >
> > I am interested in passing RDF in the metadata element of an
> > OAI GetRecords
> > response so when I write a harvesting application and can
> > pass the content
> > of the metadata element off to an RDF storage tool (like
> > Redland, RDFStore,
> > or rdfdb) without further processing.
>
> You are going to have to do some processing anyway, right?  You'll have
> to pull out the metadata package from the larger OAI protocol response.
> The step of "turning that into RDF", wrapping it in the RDF tags, is
> minimal.
>
> I think the idea of exploiting RDF is sensible, I don't think that
> embedding the XML data in an OAI response in RDF tags is a great idea.
>
> Let me be more specific.  In OAI we have essentially established a
> protocol for the Warwick Framework concept, explained in
> http://cs-tr.cs.cornell.edu/Dienst/UI/2.0/Describe/ncstrl.cornell/TR96-1
> 593?abstract= - making distinct packages of metadata available.  What we
> haven't done is dealt at all with the relationships among those multiple
> packages - e.g., what does a MARC xxx field have to do with a DC foobar
> element.  IMHO, this is an issue better left to the service level rather
> than the provider leve.  Why?  1) Because there are possibly multiple
> intepretations of such relationships 2) Because understanding and
> expressing such ontological thingies is usually not the area of
> expertise of the average archive/repository manager.  OAI is targeted at
> the kind of folks who usually don't dwell at that level.
>
> Now, I think that it is entirely reasonable to create a metadata
> integrator service.  Such a service might devise a set of RDF schema (or
> other mechanism) that express metadata vocabulary interrelationships.
> That service could then harvest different metadata packages (in diff.
> vocabs.) from data providers and populate a database of canonicalized
> metadata, that could then be expressed in other formats or vocabularies
> (again derived via something like RDF schema).
>
> This is actually the kind of thing we have been playing with in our
> Harmony project http://www.ilrt.bris.ac.uk/discovery/harmony/ and in a
> metadata model called ABC.
>
> > For example, it seems possible for me to convert the entire
> > corpus of the
> > Open Directory Project into RDF. I could then save this data
> > into some sort
> > of database application such as Redland, RDFStore, or rdfdb.
> > Once in one of
> > these sorts of applications I can provide searching and
> > reporting mechanisms
> > against them. I could then use OAI to harvest the content of
> > the "deep Web"
> > -- the content of databases, have the metadata returned in
> > RDF, and then
> > save this data to Redland, RDFStore, or rdfdb as well. OAI
> > strengths seems
> > to be the provision of an API for querying remote resources for their
> > metadata. RDF's strength lies in describing how that metadata
> > is structured.
> > Why not combine them?
>
> RDF doesn't describe how "metadata is structured".  It merely provides a
> set of primitives for modeling resource relationships and types.  As
> said above, its not that I think that RDF is a bad idea - I think it
> makes great sense.  However, it may be more appropriate at a higher
> level (service level) than what we've defined OAI for.
>
> Carl
>
> >
> > More to the point, I believe I am more interested in #1, #3,
> > and #4 above. I
> > would like to leverage the ability to mix and enhance Dublin
> > Core tags, akin
> > to the use of exploiting RDF primitives, and I would like to expose my
> > metadata in RDF for further processing.
> >
> > --
> > Eric Lease Morgan
> > Digital Library Initiatives, NCSU Libraries
> > http://www.lib.ncsu.edu/staff/morgan/
> >
> >
> > _______________________________________________
> > OAI-implementers mailing list
> > OAI-implementers@oaisrv.nsdl.cornell.edu
> > http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-implementers
> >
> _______________________________________________
> OAI-implementers mailing list
> OAI-implementers@oaisrv.nsdl.cornell.edu
> http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-implementers
>