home · mobile · calendar · colloquia · 1998-1999 · 

Colloquium - Craven

Combining Relational and Statistical Methods for Information Extraction
Carnegie-Mellon University

For many applications, information in text form remains a greatly underutilized resource. In a pair of related projects at CMU, I have been developing new methods that enable information represented in text and hypertext sources to be used as if it were in a structured representation, such as a knowledge base or a database. In the WebKB project we are developing methods for automatically constructing knowledge bases by extracting information from the Web. In the BioKB project, we are developing methods for extracting knowledge-base facts from on-line biomedical text sources, such as MEDLINE.

One focus of my research has been to develop a new learning approach that combines a relational learning algorithm with a statistical text classification method. This algorithm often induces more accurate classifiers than either a purely statistical method or a purely relational one. I will describe this approach and discuss its application to extracting information from both the Web and biomedical abstracts.

Hosted by James Martin.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)