Automatic chinese sentiment word extraction based on maximum entropy

Abstract

In recent years, sentiment analysis has become a hot topic in the study of natural language processing. Methods of machine learning are widely used to the sentiment analysis. This paper presents an approach for Chinese sentiment analysis at phrase-level. A LMR template is designed to tag word features, like position, orientation, part of speech (POS), and so on. Then, Maximum Entropy (ME) model is employed to extract sentiment words. Parts of the first Chinese Opinion Analysis Evaluation (COAE2008) corpus are used in evaluation. Experimental results show that ME model with LMR template can achieve a good performance. © 2009 IEEE.