Paper: MIEA: a Mutual Iterative Enhancement Approach for Cross-Domain Sentiment Classification

ACL ID C10-2152
Title MIEA: a Mutual Iterative Enhancement Approach for Cross-Domain Sentiment Classification
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010
Authors

Recent years have witnessed a large body of research works on cross-domain sentiment classification problem, where most of the re- search endeavors were based on a supervised learning strategy which builds models from only the labeled documents or only the labeled sentiment words. Unfortunately, such kind of supervised learning method usually fails to uncover the full knowledge between docu- ments and sentiment words. Taking account of this limitation, in this paper, we propose an it- erative reinforcement learning approach for cross-domain sentiment classification by si- multaneously utilizing documents and words from both source domain and target domain. Our new method can make full use of the rein- forcement between documents and words by fusing four kinds of relationships ...