Research Of The Chinese Text Automatic Classification Based On RBF Network

With the rapid development of World Wide Web , the number of information from the web has increased at high index speed. It has become more and more important and difficult for people to arrange the large amounts of text quickly and effectively by classification .So studying the text automatic classification has very important and real meaning.After studying the automatic text classification, this paper has found artificial neural network has strong self-learning, self-organizing, associative memorizing and reasoning functions and has special advantages on the automatic text classification. The artificial neural network which is now often applied into the automatic text classification is BP(Back Propagate) network, but its disadvantage is local minimization and slow in convergence, which has influenced the effect of the automatic classification. The operational speed of RBF (Radial Basis Function) network is faster than BP network. Especially RBF has strong nonlinear mapping ability and has been applied widely in many areas.Based on the above reasons, the author has designed and implemented the Chinese text automatic classification based on RBF network. The main designing idea is as follows. First, take some text randomly from the text corpus as training text. Second, set up a segmentation dictionary for word segmentation and delete the stop words and punctuation and then count the words. Third, calculate the mutual information of each word and extract some words as the feature words from each classification. Here the mutual information means the relevance between word and classification. And then code every feature word and text to make the text recogonized by the computer. In the end, build the RBF network text classification and initiate the network and use the training text to train the RBF network until the whole network satisfies the convergencecondition. Now the automatic classification of Chinese text based on RBF has been set up.We have obtained pretty good results using this Chinese text automatic classification based on RBF. The classification precision can always keeps about 85% at different radial basis radius. In a word, studying the application effect and potential of the Chinese text automatic classification based on RBF, we can find out that RBF are better than BP on the aspect of approximation ability,learning speed and classification ability .So this is a good method of automatic text classification that is worth of further study.