Date of Award

Spring 1-1-2011

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

James H. Martin

Second Advisor

Michael C. Mozer

Third Advisor

Rodney D. Nielsen

Abstract

Relation extraction is a means of determining relations between entities discussed in text. In this work we apply a number of existing and novel techniques to the J.D. Power and Associates Sentiment Corpus to examine the techniques' effectiveness when looking for relations in noisy blog and social media documents. To classify relations, we use support vector machines with both vector features and tree kernels, the current state of the art methods for relation extraction. Additionally, we extend these methods to examine relations across multiple sentences. Unlike previous systems, this one focuses predominantly on relations between all entities discussed in the text rather than only entities that are mentioned in the same sentence. We present evidence that relation extraction can be improved using an ensemble classification scheme to combine relations between mentions to predict relations between co-reference groups of mentions.

Share

COinS