I attended two tutorials at ACL. The first was in the morning, and I happened to present part of one of them (with Rada Michalcea). It was called "Advances in Word Sense Disambiguation", and I think it was pretty good. :) You can find the slides here:
http://www.d.umn.edu/~tpederse/ADVANCES-IN-WSD-ACL-2005.ppt The idea behind our tutorial was to give an overview of WSD methods, including knowledge-based, supervised, unsupervised, and semi-supervised. The second was called : "Max-Margin Methods for NLP: Estimation, Structure, and Applications" and it was presented by Dan Klein and Ben Taskar. Slides are available here: http://www.cs.berkeley.edu/~taskar/pubs/max-margin-acl05-tutorial.pdf This was quite a bit more mathematical, and generally was making a case for the use of discriminative models over generative models. The difference is that generative models are usually making assumptions about the form of a model, and are usually pretty easy to estimate, assuming you have adequate data and so forth. In the end a generative model captures a joint probability estimate, as in p(A,B,C,D). A discriminative model is not always (or ever perhaps?) based on a probability distribution. The best and simplest example of a discriminative model is the perceptron. Support Vector Machines (SVMs) are the currently most popular discriminative model I think, and so in some respects I viewed this tutorial as the "SVM tutorial", although that isn't technically correct, the focus was on max-margin methods, of which SVMs are one instance. This tutorial can be viewed as an extension in some way of an earlier tutorial by Dan Klein and Chris Manning, which is summarized here: http://groups.yahoo.com/group/nlpatumd/message/211 Finally, you can still get the slides of that earlier tutorial here: http://www.cs.berkeley.edu/~klein/papers/maxent-tutorial-slides-6.pdf Enjoy, Ted -- Ted Pedersen http://www.d.umn.edu/~tpederse 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/
