一种基于频次统计的兼类噪声消除方法  

An Algorithm to Reduce Multi-Category Noise Based on Frequency Statistics

作　　者：尹中航[1] 王永成[1] 宋聚平[1] 蔡巍[1] 

机构地区：[1]上海交通大学电子信息学院,上海200030

出　　处：《上海交通大学学报》2003年第3期408-410,共3页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目 ( 60 0 82 0 0 3 )

摘　　要：分析了自动分类知识库中的文本兼类噪声 ,提出借助于频次统计特性来减少兼类噪声的新算法 .在进行理论分析的基础上 ,讨论了具体的实现步骤 ,并通过对新闻语料的分类实验 ,检验了降噪效果 .结果表明 ,该方法可以减少兼类概念在知识库中的冗余次数 。This paper analyzed the multi category noise in knowledge base for automatic text classification, and presented a new algorithm by means of the statistical characteristic of knowledge base. Based on the theoretical analysis, it discussed the concrete steps of the algorithm, and tested the effect by classifying news samples. The experimental result indicates that the algorithm can obviously decrease the redundant appearing times of concepts of multi-category samples in knowledge base. The performance of automatic classification was improved after revising.

关 键 词：知识库 降噪 自然语言处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]