I have just come back from Singapore, from the DC-2007 conference (the annual conference of the Dublin Core Metadata Initiative). It was a good week; I learned a lot…
The DCMI “bridge”
One thing I realized we tend to forget is that DCMI is more than those “dc” terms that we all know and use. DCMI has a number of working parties developing different vocabularies based either on the classical “dc” terms (or their new incarnation) or on new terms. Topics include a vocabulary for accessibility or working with IEEE on a version of the Learning Object Metadata (LOM) (essentially, a vocabulary geared towards e-learning). And because DCMI has a clear commitment to RDF (see, eg, the document on the DCMI abstract model and its mapping on RDF), this means development of vocabularies that enrich the Semantic Web, too!
In fact, DCMI also plays a very important role of a bridge between the Semantic Web and the traditional librarians’ worlds. Having contacts with the community of librarians is really important and, frankly, I think until now that community was more scared from the Semantic Web rather than anything else. The SW seems overly complex and complicated for them and they do not yet see the benefits of adopting it. In this sense, this conference was also good: there were a number of presentations referring to Semantic Web usage; hopefully, my tutorial also contributed a bit. Something to follow up, that is for sure.
RDA (Resource Description and Access)
Maybe the best example for this bridge role is the work that DCMI has started in cooperation with RDA experts. I now begin to understand a little bit what RDA is all about. The short story is that it aims at becoming the standard cataloging rule set to be used by all major libraries around the globe, whether university libraries or other, public institutions (eg, Library of Congress in the US). Well, a working party has been set up a short while ago to develop a version of RDA along the DCMI style of work, ie, developing that standard into DCMI style vocabulary. Hence, automatically, RDF, too! What this could mean is that all major library catalogs around the globe could be exposed in RDF, all using the same vocabulary, and could therefore be merged with, say, dbpedia or other data on the Semantic Web. It will take a few years, but we are talking about a huge, sorry, HUGE amount of data that would be linked to the Semantic Web… it is an exciting prospect.
Another, though very different, area of interesting discussion was what the DCMI experts call “application profiles”. Here is the problem. You have some vocabulary, described in RDFS. You then have a special user (sub-)community that wants to use that vocabulary but with some strings attached (eg, restrict the cardinality of a specific predicate, or the value it can take). The set of rules (or maybe we should call them “conventions”?) that are used to describe those restrictions are what DCMI call application profiles. (I am a bit vague, but the notion is still under development, no fixed specification yet.)
If something like OWL was used for that vocabulary, than one could think of using the property restriction mechanisms of OWL to achieve something like that. Of course, that would mean defining a set of properties in a new namespace (using subproperties) and then restrict those, but that may be all right. The problem is that the current DCMI vocabularies are not described in OWL and, I guess, OWL would be considered as a sledge hammer by that community. Hence the perceived need to find a different, sort of orthogonal mechanism to property restrictions for simpler vocabularies. The current line of thoughts is to define, essentially, a set of rules that describe the constraints for the RDF graphs that would abide to a specific application profile. By “rule” I do not necessarily mean formal (eg, Horn) rules (although that may be a valid approach, who knows?). The idea of using SPARQL as some sort of a “checking” engine did come up in the discussions: one would define a number of SPARQL queries with either ASK or CONSTRUCT, and the result of those queries could be evaluated to decide whether a specific graph is ok or not. (Actually, to come back to the usage of OWL: if the tractable fragments work leads to some standards eventually, that may change the “sledge hammer” perception…) All this is still in the making, but worth keeping an eye on!
One last issue to remember and to note. I learned that DCMI intends to fully incorporate GRDDL, in the sense that:
- a profile document will be defined that can be used if an author adds DC metadata in the header section of an HTML file
- the profile document will lead to a proper transformation to generate the “right” RDF data
This is not yet in place but should relatively soon…
As I said: it was a good week.