基于支持向量机的汉语问句分类

Chinese Question Classification Based on Support Vector Machine

目前汉语问句分类一般都依据疑问词及其相关词的组合规则,但由于规则的提取很深地依赖于语言知识,而且很难穷举出所有的特征规则,因此会影响分类的效果.支持向量机(SVM)是建立在统计理论基础上的机器学习方法,对于小样本分类问题有很好的识别效果.文中分析和定义了汉语问句的类型,建立了以SVM为基础的问句分类模型,详细描述了问句分类特征的选取过程,并在句法特征的基础上引入语义特征进行汉语问句分类实验,分类准确率达88.7%,表明结合句法和语义特征以SVM进行汉语问句分类具有很好的效果.

At present, Chinese question classification is commonly based on the combinatorial rules between the interrogatives and their interrelated words. Because the extraction of the combinatorial rules greatly depends on language knowledge and not all combinatorial rules can be listed, the classification performance is not desirable. As the SVM （Support Vector Machine）, a machine learning method based on the statistical theory, possesses excellent discriminating effect on small sample classification, this paper establishes a question classification model based on SVM after the analysis and definition of Chinese question types. The process of the feature selection for question classification is then described in detail. Finally, a question classification experiment is carried out by introducing corresponding semantic features based on syntactic features, with a classification accuracy of 88.7% being achieved, which indicates that Chinese questions can be excellently classified by means of SVM with the combination of syntactic features and semantic features.