关键词: 网络新闻评论, 中文信息处理, 情感分类, 机器学习, 口语化评论

Abstract: Netnews comments has become an important channel to express personal opinions for the common people, and sentiment analysis can find out the whole attitude of the common people for the news events. This paper summarized the characteristics of netnews comments firstly, and selected different sets of feature, different feature dimensions, different feature-weight methods and parts of speech to construct classifiers; then made the comparison and analysis to the experimental results. The results of comparison show that the features combining sentiment words and argument words perform well to those only employing sentiment words; otherwise, feature dimension has less influence on the accuracy of classification for this kind of data, and the feature-weight method of TF-IDF is still better than boolean method. As for part of speech selection, nouns and verbs as features obtain better performance than adjectives and adverbs.

Key words: netnews comments, Chinese information processing, sentiment analysis, machine learning, oral comments