关键词: 情感极性, 网络新闻评论, 评论关系, 扩展情感词典, 支持向量机

Abstract: Sentiment polarity analysis of Web news comments will be of great importance in analyzing public opinion and mastering the polls in the Internet age. Nowadays researches are focused on analyzing the comment itself and ignoring the structural?relationships between the comments, so the paper uses the relationships to build comments relationship tree, and constructs the sentiment polarity decision?rules with the tree. After pre-processing, it takes the both methods based on sentiment dictionary and  on SVM to analyze the sentiment polarity, extends the sentiment dictionary dynamically and designs the classifier. Experiments demonstrate that the methods get higher precision after the comments relationship tree is bringing in.

Key words: sentiment polarity, Web news comments;comments structure, extension emotion dictionary, Support Vector , Machines（SVM）