Appraisal Expression Recognition Based on Generalized Mutual Information

The polarity of a subjective sentence in customers’ reviews depends on not only the semantic orientation of opinion words, but also their context, especially the modified features. An Appraisal Expression consists of an opinion word and the modified feature, and can accurately describe the holders’ opinion about the feature. In this paper, we propose a novel method that uses generalized mutual information to automatically recognize the appraisal expressions from customers’ reviews. Our method does not fill in any template and is domain independent. More important, our method can avoid the complex syntactic analysis while keeping the comparable accuracy, which greatly improves the efficiency of appraisal expression recognition. Our experimental results show that the F-measure of our method is up to 80.18% and 80.08% respectively, which is higher than the existing methods, and the efficiency is also comparable to the past methods.