Virtual Communities Adverse Information Filtering Technology

With the development of computer technology, the network virtual community is appearing, the activity of network public opinion has reached unprecedented level. But many bad network information and reactionary comment also is appearing, the social spiritual civilization has been polluted, and safety of the country has been damaged. The bad network information must be filtered, and the network environment must be purified now.According to the bad information feature of network virtual community, the author researches the word feature, structural feature, system model and so on. The filtering model is implemented by Bayes method and SVM method for the bad information of network virtual community.In this dissertation, the author first is working to deal with special symbol or unusual words for word processing. According to the feature of the texts that from the network virtual community, author creates the stop words table and bad words table, and use the method of DF to get 250 feature words.Second, author analyses the feature of many text categorization methods, filtering model of the network virtual community is implemented by Bayes and SVM. The experimental result by SVM is better than Bayes about precision, recall and computing speed. The result by SVM is 92%,98.7%,95.2%.Finally, the author designs and implements the prototype system to filter bad text that from the network virtual community by J2EE.