| Text classification is a hot research topic and core technology in the field of information retrieval and data mining at present. It has a very important application on the web page classification, topic classification of Press and Publication, personalized news, spam filtering, users'categorization on SNS, and user classification on micro-blogging and so on. Therefore, it has a very important practical value to do a depth research in text classification.First of all, in this paper, the currently study status of the text classification technology at home and abroad is introduced. And then the relevant text classification techniques, which mainly include some commonly algorithms of text classification and some popular methods of feature selection, is discussed. Each of the algorithms is conducted in detail. Meanwhile, each method in the situation where it is applicable and not applicable is explored. Besides, the advantages and disadvantages of each method also are listed out. Secondly, this paper elaborates the concept of sense group and its related methods and deduces the feasibility of building the category vocabulary by the sense group and its related methods. After that, it explores how to establish the classifier of text categorization by the category vocabulary, which including the text classifier of SVM based on sense group and the text classifier of Naive Bayes based on sense group. So, the text documents are classified by the text classifier. Finally, we test the methods which are proposed previously by experiment. The results showed that the text classifier of SVM based on sense group is better than normal SVM not only in the speed of text categorization but also in the accuracy and the recall, but the one of Naive Bayes based on sense group is worse than Naive Bayes in performance. Therefore, when the text document is needed to classify, if the classification algorithm can be appropriate chosen, the text classification based on sense group will achieve a good result in performance.Based on some in-depth study of common text classification algorithm, the new method of text categorization is proposed by this paper. At the same time, it is strictly deduced in detail, which will play a promoting role in the field of text categorization. |
