Design And Implementation On The Text Classifier Based On Support Vector Machine

With the popularity of the Internet and e-commerce applications extensively, people enjoy surfing the Internet, but also caught in the dilemma of information overload. It is difficult for users to find something useful in a lot of information. Businessman is also difficult to master network on product review information. As a result, the text classification system came into being.Text classification system is an effective tool in dealing with massive text. It can provide a good set for the text's organizational structure, so as to reduce the scope of the text search and improve the efficiency of text processing. Text classification system has a wide range of applications in digital libraries, e-mail classification, news categories, text retrieval and other fields. For example, news text classification needs to be divided into economic, political, sports, entertainment and the like, to facilitate users to browse; e-mail messages need to be classified into spam and non-spam to the spam filter. Text classifier is an important part of text classification system, with good prospects for the development and application. It has become an important research content.The text classifier based on support vector machine is the focus of current research. It shows outstanding advantages in many ways than other classifiers. However, the text classifier, based on SVM, is not mature enough. There are still some problems, such as system scalability issues, ease of use and the time factor issues. This article is mainly to the research on the text classifier, which based on support vector machine.In this paper, through literature, we summarize existing characteristics of the text classifier and then introduce the support vector machine and the basic theory of text classification. So we have a clear understanding on the text classifier. The last is the text classifier's design and implementation. The text classifier consists mainly of three parts:1,Text shrinkingAfter word segmentation and syntactic analysis of text messages, we get to do some simple reduction process. So we can improve results of the text training and text prediction. The purpose of the text data reduction is that:â‘ This can avoid some of the characteristics of range, which is too big and some other features too small.â‘¡This prevents numerical difficulties when computing the inner product. So you can calculate the kernel function better.2,Text trainingThe main purpose of the text training is to construct the text classifier. Through the study of the given relevant information about the classification system, we can use the weight of feature items, calculated by the algorithm of feature weights, to construct the text classifier.3,Text predictionThe key of text prediction is to build a text classifier, which must be accurately classified. There are many text classifiers, which achieve very good results in different areas. How to evaluate the performance of a classifier more objectively? It is worthy of research.In addition, this article carried out evaluation and analysis on the text classifier through experiment in the final.