Hi friends,

Just thought of writing about my new research here at Pitt.
It fits, builds quite nicely upon my previous work with
SenseClusters and on unsupervised text clustering.

I am working with Dr. Diane Litman on a dialogue based physics
tutor, itspoke. The specific problem I am working on is to

1. detect various topics (physics concepts like newton's laws
of motion, freefall, vectors) that the tutor covers and then,

2. find how it makes transitions between those topics,
e.g. in what order it covers the necessary concepts like "newton's
laws of gravitation", "F=m*a", "air resistence" etc before teaching
to solve the problem of "freefalling objects" and so on !

It is very much based on the famous "catching the drift" paper by
Regina and Lillian. The main difference is that, we deal with
dialogues which have quite unique features from normal text
articles (monologues) that they used.

Once we find such topic transitions, we can then see how planned
the tutor is and how closely its actions (sequence of topics)
follow the plan specificied by developers, or maybe to see how it
correlates with a human tutor teaching same concepts.

For this, the first step is to create clusters of dialogues
that reflect the various concepts/topics the tutor covers,
and then create the discourse model that shows the transitions
between those topics/clusters... No need to say, I use my favorite
old little SenseClusters for clustering dialogues :-) Its quite
nice to use the same package for a entirely different dataset
with a different perspective.

I am going to talk about this very soon in our group meeting,
so thought would be nice to collect some feedback from you all,
since you all know SenseClusters quite well. So, feel free to
send me your comments, questions, suggestions etc ...

Thanks,
Amruta

___________________________________________________________________
Amruta Purandare                    [EMAIL PROTECTED]
Intelligent Systems Program         http://www.cs.pitt.edu/~amruta
University of Pittsburgh            (412)-657-1318
___________________________________________________________________



------------------------ 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/
 



Reply via email to