WWW 2006 - Recap of the past two days
So two days have past, so I thought I’d catch you all up to speed on whats been going on. First I’ll talk about the happenings two days ago, then talk about yesterday.
So, two days ago (5/23/06), the conference was officially open. The opening ceremony was actually quite fun…there were bag pipe players and traditional Scottish dancers. I’ll post some pics once I can get them uploaded from my camera. The opening keynote was given by David Brown, the chairman of Motorola. His talk was focused on the transition to a mobile marketplace and was quite interesting. In particular he expressed the view that seamless mobility between content handling and passing is where the future is heading. Additionally he claimed that people value continuity of experience and that further personalization in the mobile arena is needed. In general his insights seemed on point. One thing I especially found interesting was his discussion about the current efforts to bridge the digital divide – that is the lack of (digital) infrastructure in developing areas. What was interesting was that there are currently plans/efforts to develop low cost handsets (with an intial base cost of $15-$30) to provide to developing areas; further there are plans to work with local governments to reduce taxes on such devices.
I also attended the IPTV workshop, which addressed a variety of issues with respect to streaming TV over IP. The talks mainly focused around frameworks for delivering TV of IP in a more effective manner. One of the more interesting talks detailed a new/extended P2P morphology for delivery of TV streams. Their approach seemed ok, however their UI (which the claimed was superior) left me wanting more.
During the afternoon session I attended the Identity, Reference, and the Web workshop. The main goal of the workshop was to address the various views of how to identify (syntactically, semantically, philosophically, etc) resources on the Semantic Web. It was interesting to observe all of the differing views, including no http, only http, deprecate all prior naming schemes and use PRIs, etc. etc.; it was even more interesting to observe how heated everyone got when discussing the issue. It’s clear that everyone has their own needs, requirements, and preferences (for whatever their reasons may be, be it political, practical etc). During the workshop, many good arguments were provided. The main problem is going to be getting the community as a whole to jointly on a solution…this will take a long long time (probably years) to resolve (if its even possible).
Yesterday’s session opened with a technical panel discussion that included Nigel Shadbolt, Jim, Tim Berners-Lee, Clare Hart (from Factiva), and Richard Benjamins (iSOCO). The theme of the panel was ‘The Next Version of the Web’ … the Semantic Web. The panel was interesting, however most of the audiences questions were what is typically asked. This included ‘Whats the killer app of the SemWeb’, ‘Whats the return on investments of the SemWeb’, etc. etc. One of the most interesting questions/responses was actually by Jim. The question was something along the lines of ‘What is the biggest surprise in the progress of the SemWeb, now that you look back a few years.’ Jim’s response was that when they wrote the Scientific American article they largely underestimated the scale…that we now have to think about 10^10.
During the day, I attended the Ontology track which had some interesting papers. First Peter Patel-Schneider presented his and Ian Horrocks’ view that the classical logic paradigm (for their purposes DL) should be used for the Semantic Web rather than the datalog paradigm. The talk was very good and he provided various arguments against the CWA, UNA etc of datalog. The next talk addressed the topic of ontology segmentation. While the work is well motivated (large/complex ontologies are problematic for reasoning, for example say to classify, so lets reduce their size somehow), however the approach was, in my opinion, far from any usable solution. Essentially the authors provide an algorithm for selecting a subset of the concepts and relations from the ontology to be considered for reasoning. Among many other things, the authors propose traversing up the (asserted) class hierarchy and only consider classes reachable up the subsumption tree (again there are other methods for adding/reducing additional concepts from the segment). But here’s one of the many problems. As we haven’t classified the ontology, we don’t even know the real concept hierarchy, therefore the they can’t make any guarantees that any of the inferences in the original ontology will be guaranteed to be maintained after the ontology is segmented. So, we end up with a useless segment. I mean, who cares if we can chop up an ontology into bits and pieces which have no semantic correspondence to the whole. Anyway I could rant more, but I’ll save my breath.
Later in the afternoon I attended the Semantic Tagging track and I’ll briefly mention some of the work that Yahoo presented that was very interesting. As we all know they’ve recently taken over Flikr, so Andrew Tomkins presented their work on visualizing Flikr tags as their use evolves over time. In the paper, they present an efficient algorithm to process streams of tag use information and mine the representative tag (that used most in a given time interval) image from the stream. The algorithm they presented was interesting, and even further their implementation was beautiful. Hopefully they will merge this service into Flikr itself.
Anyway more to come soon…
