Question-answering system can answer the question described in natural language by using some technologies of natural language processing, information retrieval etc. It plays an important role in Distance education. To increase matching speed, there should be an appropriate classification and index techniques. This paper is devoted to resources classification in Question-answering system through applying the SVM and ME model. The main contents are as follows.1. The fundamental theory of classification was introduced, and the key technologies, such as Chinese word segmentation, Feature selection, weights computing, the mathematical fundament of model theory and algorithms for parameter estimate, were discussed.2. In the pretreatment process of resources, we got the contents from complex format documents by using a series of opening source software. We also developed a full-cut Chinese word segmentation method.3. We achieved a multi-classification supporting vector mechanics by using acyclic graph structure and estimated the parameter by using SMO algorithm. And then we analyzed the result of the classification when the model adopted different feature selection methods and different kernel functions.4. In the implementation of classification system which uses ME mode, the parameter was estimated by CGGIS algorithm. For a binary-function on events, we took the word frequency and log (frequency) as the function's value, and analyzed the result which was generated in these three conditions. |