A ME Model Based on Feature Template for Chinese Text Categorization

Abstract- With entering into information society and the Internet developing rapidly, people could acquire more and more information. How to utilize Internet information efficiently and promptly, has became a hotspot in information technology. Text categorization is an important component to help getting useful message from tremendous amount of vast information. And it assigns new documents to pre-defined categories automatically based on their content. In this paper, we puts forward propose an improving approach to apply ME model to Chinese text categorization, and its main points include constituting feature functions and pre-disposing documents. In order to make the feature items more reasonable, each document is undergone four steps, namely removing html tags, word segmentation, filtering empty words and removing stop words. In addition, feature-template is brought forward, which is combined with feature item’s weight while constituting feature function. The results of experiments show that the performance with ME model based upon the combination of feature-template and weight outperforms that with ME model based upon word-frequency, and both of their performance is better than that of Naïve Bayes and KNN