关键词: 垃圾短信, 贝叶斯分类, 分词, 概率, 提取

Abstract: As the progress of the spam message filtering technology, characteristics of spam message are changing all the time. Of them, spam message disguised with homonym can easily escape from filtering system. Feature that homonym shares same pinyin makes it possible that by replacing key words with pinyin it can pick up common vector and disguised vector. Making such two vectors as input of the filter system based on Bayesian respectively, it can get two independent outputs, by analyzing the outputs, the system can tell the spam message from the normal. Experimental result confirms that this system can identify spam message disguised with homonym effectively.

Key words: spam message, Bayesian classification, words spit, possibility, extract