ISWC Talks
by Taowei Wang
So the past two days have been very busy. Lots of talks went on. The IBM folks’ A-Box summary talk went well, and was nominated for Best Paper Award. The paper on SPARQL semantics is also nominated for Best Papaer Award. Yours truly gave two talks. One on Tuesday (A Survey of Web Ontology Landscape [slides],[paper]) and one today (CropCircles: Topology-Sensitive Visualization for OWL Class Hierarchies [slides],[paper]).
The first talk went OK. My colleagues thought it went well, I thought it could have been better. It was held in the auditorium where the keynote was given. Good thing not all 400 seats were filled. It was fairly well-received. One enthusiastic audience commented that the area needs this kind of survey, and he was very happy to see us taking the initiative and the painful steps to perform the analyses. One question from the audience is whether the ontologies in the survey was representative of the ontologies ‘in/for production system’. His argument was that if these are not representative samples, the survey is not useful (for him, at least). While that is a concern, for our purposes — which is to identify characteristics of ‘ontologies in the wild’ so we can design tools for the better –, those who write/read ontologies will need support, whether these ontologies are for production systems or for pedagogical. Swoogle people wree also interested, particularly that they also performed analyses on a large set of semantic documents: Characterizing the Semantic Web on the Web [paper]. They looked more into the instances on the SemWeb and characterized where they are from (geographical/business/educational domains), and what they are (FOAF etc.). Interesting graphs illustrate the powerful distributions of these files (which isn’t surprising, but certainly nice to confirm). They also confirmed (as our survey hints), in particular, that OWL ontologies tend to have more uses for ObjectProperty as opposed to DatatypeProperty. There may be a way where we can combine our efforts together to give the community a more unified way to look at the ontologies in the wild. Sir Tim was in the audience, but he didn’t ask any questions regarding how the survey impacts the tabulator 
The second talk, which is held today, went really well (at least I thought 
). The room was fairly packed (around 80 or so people). The chair (Marta Sabou) thought it was a fluid presentation of honest assessments of the tools. Large number of people have realized the importance of user-centered view of Sem Web technologies. People were very interested in the evaluation (how to avoid certain pitfalls, user samples, task-choice etc.) as we expected. Thorsten Liebig of the OntoTrack fame asked about whether having labels in visualization such as SpaceTree gave it disadvantage. We thought, and demonstrated through our experiments, that the having labels in-vis is what made SpaceTree so useful in navigation tasks! In general the audience was very interested in learning more about ontology visualization. The lack of rigorous study in the field and that the lack of understanding of what the real tasks ontologists do make advancing the field and designing useful tools very difficult. I hope to find these answers by preforming interviews and surveys of interested ontologists (there are quite a few) in the near future.
On the other note, poster sessions went well. Even PaperPuppy had plenty of visitors.
So today ISWC winds down. But Tomorrow OWLed begins.
Cheers,
Taowei
