Today I attended a talk by Guus Schreiber, co-chair of the W3C’s Semantic Web Best Practices and Deployment Working Group, at Berliner XML-Tage 2004. Some choice quotes:
On good vs. bad ontologies: “Good ontologies are used in applications. They represent some form of consensus in a community.”
“Creating my own ontology is a misappropriation of the term. Ontology is about shared understanding.”
Showing a slide with lots of RDF/XML: “You see something like this in many presentations, usually with a much smaller font … RDF/XML is a good exchange format, but not good for communicating examples.”
He pointed to some alternatives: Dave Beckett’s N3/Turtle syntax and OWL abstract syntax. A UML Profile for OWL is in the works at the OMG.
Several new Semantic Web Best Practices Notes are being worked on. Upcoming subjects include: modeling Part-of relationships, numeric constraints, and qualified cardinality restrictions (A hand always has five fingers).
Other things being worked on: Best practices for vocabulary porting and management. RDF in XHTML (much discussion, no consensus. Mailing list). Converting existing ontologies and knowledge bases to OWL and RDF, beginning with WordNet.
Guus talked about some state-of-the art Semantic Web applications people have built. All apps focused on information integration and information presentation. Ontologies were used only to help with search (more general/specific terms). No reasoning, no agents.
Guus: “As a chair I have to do expectation management; that’s why I don’t like the agent scenarios.” Agents might be 8-10 years off, a standardization effort for rules 5 years.
On scalability of reasoning engines: “Description logics is useful for verifying small ontologies. For web applications, you mostly use subclass relationships and that’s it.”
“I’m very worried that the current W3C Semantic Web effort is too much text-oriented.”
Interesting talk showing that you can build useful SW applications today. The contrast between this talk, and the research efforts presented at the rest of the conference, is very revealing. Maybe this is a good thing. Semantic Web development is moving from academia into the trenches.
(More coverage of the conference in my German blog.)
Pingback: Richard Cyganiak's Weblog » [bxmlt] Guus Schreiber: From Representations to Applications
Richard,
It’s good to see that people are discussing good vs. bad ontologies. Having said that, is it a good thing to reuse parts of well known ontologies in your own ontology as opposed to simply starting from scratch? And what happens if there aren’t any suitable ontologies for your domain?
This is applies especially to cutting edge research. And if we do develop an ontology that is used in applications, just how long will it take to come up with a concensus – Dublin Core is an example of this.
Maybe we need some kind of RFC mechanism for ontologies;-) This might be suitable for active commuities – schemaweb.info has a whole load of ontologies, however, I’m not sure how much of a concensus there is on its contents. FOAF is an obvious success story and even Edd Dumbill’s DOAP vocab is getting some attention. It is interesting to note that both are being developed faster than that of DC!
Cheers,
Rowland
The important point, I think, is that ontologies and vocabularies should be created and maintained in a community process. DC and FOAF are good examples of that. DOAP also, in a way, because Edd is a very prominent figure who is in contact with lots of people from the community. I believe that failure or success of an ontology is determined by social factors more than its technical merits. But there hasn’t been much discussion about these social factors yet.
Of course, sometimes you just have to roll your own vocabulary or ontology because nothing suitable exists. But in this case there’s the question if using Semantic Web technologies has an actual benefit.
Hm… maybe we Semantic Web developers should stop focusing on technology and start getting people to talk to each other about shared vocabularies.
I think it would be a good idea for people to talk to each other more. Apart from gaining a better understanding of the problems with sharing vocabularies, open discussion will accessible to (hopefully) a large audience. The more people know and understand about the Semantic Web and the value of shared ontologies, the greater the acceptance and adoption.
An we haven’t even got to inferencing yet!
Yeah, we need widespread discussion to develop great vocabularies and ontologies. The main roadblocks, I think, are the complexity of the Semantic Web technology stack, and the lack of an intuitive tool for developing vocabs/ontologies.
I agree 100% Manipulating RDF and ontologies is hard work. I think RDF is hard work. How am I going to get my family members to use it with their webpage? “Oh, you need to load the RDF into memory and flip the switch that fires the inference rules to make suggestions about which recipe to use for making apple pie.” ;-) That’s not going to happen in the next five minutes!
And your right about the lack of intuitive tools for vocab/ontology development – Protege is getting better. A colleague of mine at Southampton just writes his in a text editor – he claims it to be faster compared to wrestling with Protege. Maybe this has more to do with the concentration on technology than thinking about how to make the Sematic Web simpler to develop and use….