中文摘要：

本文利用三种特征选择方法、两种权重计算方法、五种停用词表以及支持向量机分类器对汽车语料的文本情感类别进行了研究。实验结果表明，不同特征选择方法、权重计算以及停用词表，对文本情感分类的影响也不尽相同；除形容词、动词和副词外的其余词语作为停用词表以及不使用停用词表对情感分类作用较大，得到的分类结果比较好；总体上，采用信息增益和布尔型权重进行中文文本情感分类的效果较好。

英文摘要：

In this paper, using three kinds of feature selection methods, two kinds weighing assignment methods, the five kinds of Stoplist and SVM on text sentiment classification are studied. The experiment results indicate that the greater text sentiment classification impact depends on other corpus, excluded adjective, verb, adverb as stop words and none stop words. As a whole, for text sentiment classification, information gain is superior to other feature selection methods and Boolean type weighting is superior to frequency type weighing.