【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 ,能提高自动分类系统的性能指标更多还原

【Abstract】 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.更多还原