一种基于朴素贝叶斯的中文评论情感分类方法研究

提出一种新的基于朴素贝叶斯的中文文本情感分类方法。该方法用情感短语作为文本特征,通过情感词典与否定副词相结合,提取情感短语,通过CHI统计法设定阈值进行特征提取,再利用朴素贝叶斯分类器进行情感分类计算。对不同CHI阈值、不同语料库、以情感短语为特征和以情感词为特征进行分类实验。实验表明,以情感短语作为特征进行朴素贝叶斯分类,在不同领域的评论中均获得了较高的查准率和查全率,证明了该方法的可行性。

A method of sentiment classification for Chinese comments based on naive Bayesian

A method of sentiment classification for Chinese comments based on naive Bayesian was presented. The sentiment phrases were used as document features in the approach. The task was decomposed into three phases: the identifying sentiment phrases by sentiment dictionary and negative adverbs, the extracting features according to threshold of CHI, and the constructing sentiment classifier based on naive Bayesian. Some experiments were conducted under distinct environments, including different threshold of CHI, different feature selection, such as sentiment words or sentiment phrases, and different area of training corpuses. The experimental results showed that the classifier based on naive Bayesian working could obtain high performance.