Paper: Exploring the Use of Word Relation Features for Sentiment Classification

ACL ID C10-2153
Title Exploring the Use of Word Relation Features for Sentiment Classification
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010
Authors

Word relation features, which encode relation information between words, are supposed to be effective features for sentiment classification. However, the use of word relation features suffers from two issues. One is the sparse-data problem and the lack of generalization performance; the other is the limitation of using word relations as additional features to unigrams. To address the two issues, we propose a generalized word relation feature extraction method and an ensemble model to efficiently inte- grate unigrams and different type of word relation features. Furthermore, aimed at reducing the computation complexity, we propose two fast feature selection methods that are specially de- signed for word relation features. A range of experiments are conducted to evaluate th...