Ivan’s private site

I just came back from a conference on Semantic Web and Digital Libraries (ICSD 2007) in Bangalore, India. Incidentally, this was also the first international event on Semantic Web in India altogether.

The idea of bringing together two communities that so clearly have things to say to one another is great. Obviously, it is also difficult; communities speak different languages, have different aspirations, goals, or traditions. There were some interesting talks (hopefully the talks will be on-line soon). Johannes Keizer from the FAO (UN’s Food and Agricultural Organization) talked about how the FAO and its partners use Semantic Web techniques for a whole range of quite interesting things (publishing site of their own journal using RDFS and with queries, integrating meteorological data with agricultural information on a pretty large scale, definition of some major taxonomies in SKOS, etc). I also quite liked the presentation of Lars Svensson, from the German National Library; I will remember his remark: “Libraries begin to realize that they are not alone”… Although there were some glitches in the programme (there were some papers that clearly did not have anything to do with that “bridge building”) all in all, it was an interesting event. Hopefully, it will help some of the participants to continue work in this area. It is clear that these two communities have a lot to say to another; as an example, the library community is probably much better in defining useful and good URI schemes for identifying, say, publications, musical compositions, etc. These are usually missing for the time being

My main issue is, however, whether this conference was a one time shot or the beginning of a series. If there is no continuation, then the initiative may not have a long-lasting effect after all. And that would be a pity. It would really be good if somebody picked up the ball and turned this into some sort of a series…

A separate issue that I tried to provoke on different fora (meetings, panel, etc.) is that the Indian Semantic Web community should be much more forthcoming than it is today. It is not my first visit to India, and I know that there are lots of really good people working on Semantic Web topics, but all in isolation. Not only their voice is rarely heard on international fora, but they do not even have contacts among one another. Maybe the organizers of this event (the Documentation and Training Centre of the Indian Statistical Institute) and/or the W3C India Office will be able to pull up some sort of a local forum. That would be good…

(Lars  Svensson sent a comment on this post telling me that his remark “Libraries begin to realize that they are not alone” was actually said by Jürgen Krause, in his plenary speech at the conference, and Lars only reused it. Duly noted, apologies to Jürgen if he reads these lines…)

The DCMI has published a draft for public review for an updated version of Dublin Core Abstract Model. From a Semantic Web point of view, the most important point about the new version is to have a clear association to RDF and RDFS terms as well as an alignment of the two models. For a vocabulary that is, as far as I know, the most widely used set of terms on the Semantic Web (after the core rdf, rdfs and owl terms, that is), this is really important. Actually, some of the “core” dc terms may also have new versions soon (in different name spaces) with more clearly defined rdfs:range and rdfs:domain, new class defitions, etc.

However… it is not all that simple and that is why it is still a draft. For good reasons: the DC community is old (in Internet terms, that is), it actually predates the Semantic Web activity at W3C, for example, which means that there are subtle but important differences in the models. An example for what I have seen is DCAM’s view of property values. In DCAM, it is possible for a property to have, say, a literal value and, at the same time, a URI value as well. An example I got from Tom Baker is (using DC’s textual syntax):

DescriptionSet (
  Description (
    ResourceURI ( <http://dublincore.org/pages/home> )
      Statement (
       PropertyURI ( dc:publisher )
       ValueURI ( <http://example.org/agents/DCMI> )
        ValueString ( "Dublin Core Metadata Initiative" )
)))

Which, in my reading, translates into something like:

<http://dublincore.org/pages/home> dc:publisher [
    rdf:value "Dublin Core Metadata Initiative";
    somepredicate <http://example.org/agents/DCMI>.
].

This means that the range of the dc:publisher should be some sort of a union of a pure Literal and some other class representing this URI+value structures. This can be done with OWL, though…

There may be other differences as well. If anybody is interested in the public comments, there is a mail on swig with further details; you can also look at the mail archives of the discussion. It is an important discussion to keep an eye on!