My first two days in San Jose were mostly occupied with being a panelist at the AAAI Doctoral Consortium. This is an excellent event, where graduate students come and present their work to a group of panelists (and their fellow students) who then make comments about their particular research topic, as well as issues relating to their presentation of the work.
Here's the program, which includes a list of the panelists and the students. You can also see extended abstracts describing the research of each participating student. http://www.csc.ncsu.edu/faculty/stamant/dc/dc04.html The two students that I served as "main" panelist for were from Columbia University in New York City, both doing very nice work in NLP. Their names are Noemie Elhadad and Smaranda Muresan. Columbia is a great place for NLP, with lots of interesting projects going on. They are both involved in work that relates to the medical domain - Noemie's research concerns how to "translate" medical text into a more readable form, while Smaranda's objective is to develop techniques with which to build a medical terminological knowledge base. There was some other work that was also quite interesting and relevant to things we do here at UMD, in particular that of Sugato Basu of the University of Texas. He does work in semi-supervised clustering, and gave us several nice pointers to work that might help us figure out how to stop our own clustering techniques automatically (once they have found the "right" number of clusters). In particular, he directed us to Tibshironi's GAP statistic, which seems quite useful. I'm going to put the URL to the Tibshironi paper here, just to make sure I don't lose it. The answer to the question "how do you know how many clusters to look for" has been something that has eluded me for some time now, so I hope this or a similar technique helps. http://www-stat.stanford.edu/~tibs/ftp/gap.pdf Anyway, this is a nice example of serendipity at a conference. I wasn't really expecting to learn about clustering at the Doctoral Consortium, but as it turns out I learned quite a bit! Anyway, the Doctoral Consortium is a fun event, in that it's a fairly small number of students and panelists, and after a little while you get to know each other reasonably well. In addition, the atmosphere is more friendly and supportive than one might find in a conference presentation setting, where the crowd is much larger and sometimes not as interested in you as an individual. There is plenty of discussion after each student presentation, and I think that might be the best part of the experience for everyone. While the talks themselves are interesting, sometimes the discussion thereafter really helps you to understand what someone is doing (and why). Then, after the first day of talks, we all went out to dinner at a very nice restaurant in San Jose. :) The Doctoral Consortium happens every year at AAAI, and you should participate in this if you are a PhD student in an AI related area (like NLP). Usually you do this after you have selected a topic for your dissertation research, but before you have really started writing the dissertation. This is so you can (hopefully) benefit from the feedback you get at the event, and thereby improve your dissertation. AAAI is very generous with Doctoral Consortium participants, they usually receive a travel grant that more than covers the cost of their trip to AAAI. (This was true this year I think, and it was also true in 1997 when I was a participant.) Finally, you can see my 1997 Doctoral Consortium abstract here: http://www.d.umn.edu/~tpederse/pubs.html Look for this (only in ps format!) Knowledge Lean Word Sense Disambiguation (Pedersen) - Appears in the Proceedings of the Fourteenth National Conference on Artificial Intelligence (AAAI-97), July 27-31, 1997, Providence, RI (Doctoral Consortium) You'll see that my interests haven't evolved all that much. :) Ted -- Ted Pedersen http://www.d.umn.edu/~tpederse ------------------------ Yahoo! Groups Sponsor --------------------~--> Make a clean sweep of pop-up ads. Yahoo! Companion Toolbar. Now with Pop-Up Blocker. Get it for free! http://us.click.yahoo.com/L5YrjA/eSIIAA/yQLSAA/x3XolB/TM --------------------------------------------------------------------~-> Yahoo! Groups Links <*> To visit your group on the web, go to: http://groups.yahoo.com/group/nlpatumd/ <*> To unsubscribe from this group, send an email to: [EMAIL PROTECTED] <*> Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
