We'll have a group meeting on Thur Sept 16 at 2pm in HH 321 (the conference room). The general topic of discussion is LSA, in preparation for Peter Wiemer-Hastings' visit.
We had planned to discuss the following paper: An Introduction to LSA, by Landauer, Foltz, and Latham http://lsa.colorado.edu/papers/dp1.LSAintro.pdf However, rather than having a discussion of the article directly, I would like to take a slightly different approach. There are some very nice demo tools available at : http://lsa.colorado.edu These allow you to run various different sorts of LSA analyses. I would like each of you to run a series of experiments/examples with these demo tools, and then prepare a short presentation to show your example, and explain why it did what it did. In order to do this, you will certainly need to draw upon material from the above paper, and also from the web site information (which is quite extensive). In particular, please read the information they provide under the following links: 1st Time User What is LSA Semantic Spaces So, here is a list of who should do what sort of demo...(these will make sense when you look at the lsa web site above). matrix comparisions - Jason (use both term to term and document to document comparisons, use at least two different topic spaces) pairwise comparisons - Jason (use term to term, term to document, document to term, and document to document, use at least two different topic spaces). nearest neighbors - Prath (use both term and psuedodoc comparisons, use at least two different topic spaces) one to many comparisons - Anagha (use term to term, term to document, document to term, and document to document, use at least two different topic spaces). sentence comparison - Apurva (use at least three different topic spaces) Each of you should prepare a series of examples along the lines I describe above, and your goal should be understand and explain what is happening as best you can (based on the information from the paper and the web site). I'd like you to put your results in the form of powerpoint slides, and that you print out sufficient copies for all of us to have and follow along with. Also, for Prath and Jason, please try and connect what you are doing above with your thesis work. I think there are some strong connections in fact, and it would be worthwhile for you to draw those connections and compare and contrast your approaches with those implemented in LSA. So, for next Thursday, each of you should have the results of your experiments put into powerpoint slides, and you should be prepared to explain as best you can what happened. Make sure you understand the differences between processing terms versus documents (if that is relevant) and also be prepared to discuss the different results you may get when you use different topic spaces (why would that matter), and then also what you think of the results in an intuitive sense. (Are they reasonable, or clearly wrong, or...) Please let me know if you have any questions about the above. Please try out your particular demo tool asap and let me know if you have any questions about it, or the task in general. Ted -- Ted Pedersen http://www.d.umn.edu/~tpederse ------------------------ Yahoo! Groups Sponsor --------------------~--> $9.95 domain names from Yahoo!. Register anything. http://us.click.yahoo.com/J8kdrA/y20IAA/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/
