[Ltg] LTG Seminar [Diego Molla, Aug 30, E6A 357]

[Rolf Schwitter]

----
LTG Seminar
 - see: http://www.clt.mq.edu.au/Events/Seminars.html

Monday, Aug 30, 2004 at 11am
Macquarie Uni, E6A 357
----

Speaker: Diego Molla
Graph-based Question Answering

Sentence similarity is an important factor to determine if a sentence
contains the answer to a question: if a sentence is fairly similar to a
question then the answer is likely to be there. Sentence similarity
cannot be a "yes-no" measure, since there are many grades of similarity
between sentences and we do not want to rule out sentences that are not
exactly the same as what we expect. In this presentation I will present
recent work developed on the use of graph similarity measures. The basic
graph framework is Sowa's conceptual graphs, but the graph similarity
measures are independent from the actual graph framework. I will show
how the logical forms used by AnswerFinder can be represented as graphs,
how to compute the similarity, and (hopefully) what is the difference
with the current similarity measures that we are using in AnswerFinder.

Much of the presentation is based on:

M. Montes y Gómez, A. Gelbukh, A. López López, Ricardo Baeza-Yates.
Flexible Comparison of Conceptual Graphs. In: Mayr, H.C., Lazansky, J.,
Quirchmayr, G., Vogel, P. (Eds.),Proc. DEXA-2001, 12th International
Conference and Workshop on Database and Expert Systems Applications.
Lecture Notes in Computer Science, N 2113,Springer-Verlag, 2001, pp.
102-111.

This paper can be downloaded directly from Springer-Verlag thanks to the
Library's current trial subscription:

http://www.springerlink.com/app/home/main.asp?wasp=n97lykuhmk5j61nvtw3p

Choose "Browse Publications" and then click "L" and choose Lecture Notes
in Computer Science link. Then search for volume 2113.

If you encounter problems to download the paper, there's an HTML version
in the author's web site:

http://www.cic.ipn.mx/~gelbukh/CV/Publications/2001/DEXA-2001-Flexible.htm