Question classification plays a crucial important role in the question answering system because categorizing a given question is beneficial to identify an answer in the documents. The goal of question classification is to accurately assign labels to question based on expected answer type. Recently, many machine learning algorithms are used for question classification. However many research results show that SVM perform best in this task, because it is well known to work well for nonlinear, sparse, high dimensional problems. In this experiment, we perform the One-against-One SVM algorithm and a feature extraction method of Chinese questions to get high classification accuracy.