基于特征恢复的中文残缺文本分类研究

A Research of Incomplete Chinese Text Categorization Based on Feature Recovery

为了更好的对残缺文档进行分类,本文以基于支持向量机的文本分类方法(SVM)和卡方统计量(Chi-Square)的文本特征提取方法为背景,提出了有监督学习模式下的两种文本特征恢复算法以及在此基础上进行残缺文本分类的新方案。与传统的直接分类方案相较,该方案在分类前通过预先对文本中残缺词恢复,实现了残缺文本的部分特征恢复。实验表明,相较于传统方法,该方案在低残缺率下,对文本分类的影响不大;在高残缺率下,该方案能得到较好的分类效果。

In order to improve the efficient of incomplete text categorization, this paper takes Chinese text categorization based on support vector machine categorization method （SVM） and chi--square statistic （Chi--Square） of the text feature extraction method as research background. This paper proposes a new method of incomplete text categorization on the basis of two kinds of text features recovery algorithm under the categorization of supervised learning mode. Comparing with traditional direct categorization method, the new method achieves part of incomplete text recovery through pre--term incomplete word recovery before text categorization. Experiments show that the feature recovery, compare to traditional methods, the new method gets little effect on the text categorization at low incomplete rates and gets better categorization results at high incomplete rates.