摘要
阐述了一个基于改进向量空间模型的中文文本分类系统的设计与实现 ,包括对该系统的结构、预处理、特征提取、训练算法 ,分类算法等关键技术的介绍 .通过引入结构层次权重系数来改进文本特征项权重 ,同时提出一种新的训练算法和文本相似度域值计算方法 .实验结果证明 :该分类系统能有效地提高文本分类效果 ,开放性测试的平均准确率在 80 %以上 ,且平均查全率达到了 86 % .
A Chinese text categorization system was developed based on the improved vector space model, including the important aspects of system structure, text preprocessing, feature selection, training algorithm, and recognition algorithm. The system introduced the structure layer weight coefficient to improve the term weighting, and a new training algorithm and a way of computing text similarity threshold were described. The test result illustrated the effectiveness of the system for categorizing Chinese text. The average precision was over 80?% and the recall was 86?%.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2005年第3期53-55,共3页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家高性能计算基金资助项目 (0 0 30 3) .
关键词
文本分类
向量空间模型
特征提取
结构层次权重系数
训练算法
分类算法
text categorization
vector space model
feature selection
structure-layer weight coefficient
training algorithm
recognition algorithm
