摘要： 根据文本数据学习的特点，采用线性支持向量机（LSVM）学习算法，实现了一个中文文本自动分类系统，并对该系统进行了针对大规模真实文本的试验测试。结果发现，系统的招回率较低，而准确率较高，该文对此结果进行了分析，并提出了一种采用训练中拒识样本信息对分类器输出进行改进的方法，试验表明，该方法有效地提高了系统的性能，取得了令人满意的结果。

关键词: 文本分类, 线性支持向量机, 招回率, 准确率

Abstract: This paper implements a Chinese text categorization system using the linear support vector machine based on the particular property of learning with text data, and the tests on the true large-scale text are made.The results show that the recall is comparatively low and the precision is comparatively high. The paper analyzes the reason and proposes a method to modulate the output of the classifier by the misclassified training texts. The experiments show that the method improves the system effectively and satisfactory categorization effect has been achieved.

中图分类号: 

• TP181