[Prev][Next][Index]

[xtag-meeting]: Meeting this week: Feb 14 10:30am




This week, Libin will talk about some of his recent work on 
information extraction.

Time: 10:30am
Date: Feb 14 Thursday
Place: IRCS Fishbowl

Machine Learning in MUC-7 Template Relation Task
Libin Shen

Abstract:

The Template Relation (TR) task is an information extraction problem 
introduced in MUC-7. In this paper, we have proposed an approach to convert 
this problem into a discriminative one. As the result, the precision and 
recall of our system is comparable to the best system presented in MUC-7, 
while almost no extra annotation work is required. In our approach, we first 
use Supertagger and Light-weighted Dependency Analyzer [Srinivas 97] to do 
shallow parsing on the text. Then features are abstracted from the shallow 
parsing result. We use these features as weak learners, and combine them into 
the final one with a boosting classifier.