Emotional element extraction based on CRFs

Abstract

As the fast development of social network and electronic business, a huge number of comments are generated by users every day. Extraction of emotional elements is an important pre-task of sentiment analysis and opinion mining for comments. In this paper, we extract the emotional elements such as the opinion holder, the comment target, and the evaluation phrase, which previous researches rarely concerned about, especially in Chinese. Based on Conditional Random Fields, we label the evaluation phrase which structure is simple. Then on account of unique characteristics of grammar and syntax of Chinese, we design several rule-based methods to extract evaluation phrase which is in complex structure, as well as comment targets and opinion holders. According to the experimental results, our method improves the performance of emotional element extraction in the domain of sentiment analysis for automobile’s Chinese comments. And it also contributes greatly to our subsequent task such as sentiment analysis of social media or comments from other domains.