WWW2006 - Experiences and Episodes: Day 1 (Christian’s View)
This will be the first of many entries I write, recounting my experiences and episodes (both good and bad) at this years WWW conference. First I’ll start with a general overview of the day’s events.
Today was the first of two days of workshops. I participated in two workshops; Semantic Web Semantic Web Annotations for Multimedia (SWAMM) and Reasoning on the Web (ROW). At SWAMM I presented a poster covering our work on NL paraphrases for image browsing. Overall the work was well accepted. Two people were even expressed interest in using the NL software within some of their tools. The only negative feedback I received was from a hardcore NLP person, who had some “issues” with us calling our paraphrases true “NL”. Essentially he thought that the translation we were doing were note completely up to par with today’s NLP techniques. In a sense he is right, so I clarified that the aim of the work was to show that more user-friendly representations (transformation of RDF/XML) of annotations can be provided to users, providing them with a better experience and further insights into the actual image annotations. I’ll note that he was satisfied with my response.
At ROW, I gave a talk on our incremental classification maintenance (caching) work. The talk went quite well and again the work was well received. Peter Patel-Schneider raised the one question I’ll note. He was curious about the memory overhead that building, finding, and caching Sets of Support (just to note SOS are cached for subsumption relations; then if there is a deletion, subsumption tests can be avoided if the deleted axiom is not in the SOS). However, as Aditya shows in his work that the computational performance overhead is very minimal and the memory overhead is (by today’s standard) not bad. For example performing tracing during for Tambis (395 classes) increases run time by only 50 milliseconds and causes 3.65mb of memory overhead.
Now for the first episode…as it turns out there were lots and lots of workshops this year at WWW; so many that they had to hold two of them and a different facility than the regular conference venue. Unfortunately, SWAMM was held at the main conference site and ROW was held at the other. Even more unfortunate was that the two venues were on opposite ends of town…a good 20 minute walk apart. So, lucky me, I got to walk between them both four times. Unfortunately this definitely reduced the number of talks I was able to attend today, but hey at least I got some exercise.
Ok, back to business. One talk/discussion that I want to comment on here was a panel session during ROW; the panel included Ian Horrocks, Peter Patel-Schneider, and Ed Barkmeyer. During the talk, the panelists expressed some of their views on reasoning on the Web, and responded to various questions. One question of interest was about what really reasoning on the Web was and how is OWL-DL really being used to reason for the Web. Their responses were quite interesting. First, both Peter and Ian openly state that today’s real applications that utilize DL reasoning aren’t really reasoning on the Web; rather they are application that need reasoning and use the Web as an utilize the infrastructure of the Web/Internet. However, all of the panelist were optimistic that the as the field progress more applications centered around reasoning on/for the Web would emerge and develop further than they are now.
One last thing I want to mention is that a majority of the talks and discussion in ROW were focused on rules languages for the Web. This area seems hotter than ever, with vast amounts of interest.
Anyway, enough rambling for now…more experiences and episodes to come soon.
-Christian
