The process of text categorization involves some understanding of the content of the documents and/or some previous knowledge of the categories. For the content of the documents, we use the filtering measure for feature selection in our Chinese text categorization system. We modify the formula of TFIDF to strengthen important keywords’ weights and weaken unimportant keywords’ weights. For the knowledge of the categories, we use association rules to improve the precision of text classification and use category priority to represent the relationship between two different categories. Consequently, the experimental results show that our method can effectively not only decrease noise text but also increase the ratio of precision and recall of text categorization.
Expert Systems with Applications 35(1-2), pp.102-110