Tuesday, 7 November 2006
The first day of ISWC was intensely busy for me. I was only able to attend a single session due to the number of people I spoke with.
I had a lengthy and interesting conversation with Harry Halpin, co-chair of the W3C's GRDDL Working Group, regarding RDFA and GRDDL. I found myself representing RDFA, which I did to the best of my ability. Fortunately, I was able to recall the critical use case for RDFA over GRDDL; the requirement to support a cut-and-paste of a block of XHTML without losing information on how that block should be interpreted. We discussed at length RDFA's potential show stopper - the lack of explicit identification of RDFA content. That prohibits searching for documents which support RDFA and leaves open the question of whether one may rely on RDFA extraction from documents where there is no a priori knowledge that a source document complied with RDFA markup. I suggest that will kill RDFA in practice unless it is addressed.
Harry cheered Eric Miller's push for a "persistent URI" service for RDF identifiers, similar to the persistent URL service operated by OCLC at purl.org.
Harry and I were tutored by IBM's Chris Welty on the difference between RDF:Resource and OWL:Thing. I can never remember the difference. RDF:Resource includes its own language-specific features, OWL:Thing does not (in OWL DL), but they are the same in OWL Full. Thanks, Chris!
Chris has made a tremendous amount of progress as co-chair of the W3C's Rules Interchange Format (RIF) Working Group. The group has reportedly agreed that they will, in fact, produce a rules interchange format (hey, that was hard!), that they will define a core feature set based on positive Horne and that they will support an arbitrary number of non-Horne extensions via an extension mechanism. That level of early structure should allow the group to proceed without the factionalization that dogged the WebOnt group (producers of the OWL standards).
Years into Semantic Web development, we still need a coordinated location for good ontologies. Harry spoke to the guy who runs http://semanticweb.org/ about hosting mappings between Web 2.0 vocabularies and RDF and ontologies. The response was positive, but it still needs to happen. Unfortunately, I did not get his name.
Apparently there is still a need for a good geospatial ontology, even for simple agreed concepts like latitude and longitude. This has been a problem for years, especially within government circles and supporting organizations such as MITRE. Harry pointed me toward Harry Chen's blog entries here.
Harry and Norm Walsh have been working on an isomorphic mapping between vcard and RDF. Details are available on Norm's blog, although Harry told me there is a newer version which he promised to send me. This is a useful thing, especially as it makes use of some FOAF to make the mapping clean.
Steve Harris, of OWL Tiny fame, is now at UK identity protection startup Garlik. Garlik is currently operating in the UK only, but they are planning a US market entry next year. They ask their customers for personal information used to identify them to their banks and watch public and subscription databases to determine if others are using their identity. Naturally, this is done via an RDF graph. There certainly is a need for some kind of identity protection service in the US. USA Today ("McPaper") reported today that 8.9 million Americans (4% of the population) lose their identity each year and that is costs them an average of US$6,383.
I finally met Chimezie Ogbuji, now at Cleveland Clinic and formerly of ForeThought. He worked on the ForeSuite CMS, which uses RDF and XML databases to manage content. Interestingly, Chimezie is a fan of Daniel Krech. ForeSuite uses Daniel's rdflib! It is also using a GRDDL-like transform between XML documents to generate the RDF.
David Taowei Wang did a good job presenting A Survey of the Web Ontology Landscape (Taowei Wang, Bijan Parsia, Jim Hendler). Bijan now has a huge unkempt beard and a nineteenth century waxed mustache. He appears to be enjoying teaching at Manchester.
The most interesting paper I have seen in a while was Semantics and Complexity of SPARQL (Jorge A. Perez, Marcelo Arenas, Claudio Gutierrez). It is up for a best paper award and probably deserves it. It is great to see someone, even if not the W3C, providing a model theoretic for SPARQL. Unfortunately, the work does not yet cover entailments or bnodes, which are outstanding issues at the W3C.
I missed seeing OntoWiki - A Tool for Social, Semantic Collaboration (Sören Auer,Thomas Riechert, Sebastian Dietzold) because I was talking to Guus Schreiber, Chris Welty, Ivan Herman and Harry Halpin (again). I'll have to read it in the proceedings, though, because it looks interesting.
The poster session was well attended. Six posters from MINDSWAP were accepted, including the Semantic Web challenge entry. Unfortunately, three of them were down a long corridor in the lunch room and received very few visitors :( I was fortunate to get an excellent location with plenty of traffic and little noise from the band.
My poster, Enhancing Software Maintenance by using Semantic Web Techniques, reported on some research in progress. My purpose in submitting it to ISWC was to get initial responses to the research direction and gather some ideas for next steps. I was pleasantly surprised to receive some very positive feedback. There was a lot of interest from software engineers, especially those from the more pragmatic people, such as those from IBM, Accenture and SRI International. Software maintenance costs money and that makes a market.
There were several comments regarding the depth of my use of OWL-DL. I had created an ontology of software engineering concepts which was focused on Java for the prototype implementation. I had used OWL-DL in order to use SWOOP to ensure logical consistency. A good next step would be to represent the high-level constructs of other languages so that multi-language projects could be managed in the environment. This is necessary because different languages treat even basic concepts differently, such as the separation of abstract classes from interfaces or the existence of unimplemented method signatures. An OWL-DL ontology could readily map the like and disjoint components in the various languages and be used to infer inheritance relationships.
Dr. Kerry Taylor from CSIRO in Canberra stopped by. She knows my advisor Dave Carrington at UQ and was surprised to see him involved in SemWeb work. She confirmed the earlier comments regarding inferencing across language differences.
I must see similar project called FAMIX. FAMIX provides "a language-independent representation of object-oriented source code and is used ... as a basis for exchanging information about object-oriented software systems." Avi Bernstein of the University of Zurich told me about it and recommended that I discuss it with his collaborator, Harald Gall. This is what conferences are for.
Post a Comment