Appraisal expression extraction based on parse tree structure
基于句法树结构的情感评价单元抽取算法*

Appraisal expression extraction is a fundamental task of sentiment analysis. Most of the previous methods for appraisal extraction are based on flat features such as syntax path which suffer from the noise in Chinese reviews. This paper proposed a phrase tree structure to present the appraisal expression pattern, as well as a method based on approximate convolution tree kernels to calculate the similarity between these structures. Then it used a similarity-based pattern matching method to extract the appraisal expressions. It conducted the experiments on Chinese product reviews, compared with the baseline which used the syntax paths as features, the precision was 13.4% higher and the recall was 9.2% higher. The experiment results showe the approach is efficient for appraisal expression extraction on Chinese product reviews.