Abstract: For improving the efficiency and accuracy of Chinese text categorization，this paper presents a new Chinese text classier，in which a novel feature selection is proposed according to word frequency，mutual information and classificatory information，and after analyzing the hypostasis of the traditional TF-IDF，a weight adjustment method is put forward in which the IDF function is replaced by function used in feature selection.Finally a fast Bayes theory classier is designed.Experiments prove this classier is simple and effective.

Key words: Chinese text categorization, feature selection, feature weighting, classification algorithm

摘要： 为了提高中文文本分类的效率与精度，设计了一种新型的分类器。该分类器采用基于词频、互信息和类别信息的综合评估函数进行选择特征；在特征权重计算上，由于传统TF-IDF方法没有考虑特征类间和类内分布，提出了一种将词频和综合评估函数值相结合的权重计算方法；最后设计了一种基于贝叶斯原理的快速分类器。实验证明该分类器简单有效。

关键词: 中文文本分类, 特征选择, 特征权重, 分类算法