TARGET-DEPENDENT SENTIMENT ANALYSIS FOR PRODUCT COMMENTS

Abstract

Traditional sentiment analysis method analyzes the whole sentiment polarity of comments without concerning about the relevant targets. Existing target-dependent sentiment analysis usually ignores the multi-target and multi-opinion sentence, which causes wrong target identification. In this paper, we propose a novel target-dependent method based on Conditional Random Fields (CRFs) and syntax tree pruning. A parallel tri-training method based on MapReduce is used to label corpus semi-autonomously. CRF model is used to extract positive/negative opinions and the target of opinions from comment sentences. Syntax tree pruning is used to prune the irrelevant target of opinions and extract the correct appraisal expressions. Finally, a visual product attribute report was generated. Through extensive experiment, the accuracy of the proposed method on sentiment elements and appraisal expression can reach 89% approximately. Which shows our method outperforms traditional methods on both sentiment analysis accuracy and training performance.