• 人工智能及识别技术 •

### 基于增量式贝叶斯模型的中文问句分类研究

1. (安徽工业大学计算机科学与技术学院,安徽马鞍山243002)
• 收稿日期:2013-09-09 出版日期:2014-09-15 发布日期:2014-09-12
• 作者简介:王小林(1964 - ),男,教授,主研方向:人工智能,中文信息处理;镇丽华,硕士研究生;杨思春,副教授、博士研究生;邰伟鹏, 讲师、博士研究生;郑　啸,教授、博士。
• 基金项目:
国家自然科学基金资助项目(61003311);安徽高校省级自然科学基金资助项目(KJ2011A040)。

### Chinese Question Classification Research Based on Incremental Bayes Model

WANG Xiao-lin,ZHEN Li-hua,YANG Si-chun,TAI Wei-peng,ZHENG Xiao

1. WANG Xiao-lin,ZHEN Li-hua,YANG Si-chun,TAI Wei-peng,ZHENG Xiao
• Received:2013-09-09 Online:2014-09-15 Published:2014-09-12

Abstract: Since the performance of the classifier generated by the fixed training set is not satisfactory and can hardly track the users’ needs dynamically,in this paper,the incremental Bayes idea is introduced in question classification. In order to eliminate the feature redundancy in the training set,Genetic Algorithm(GA) is used to select the optimal features to amend the classifier. In the process of classifier learning,the parameters are modified dynamically while the training set is expanded. The interrogative word, syntax structure, question focus words, and their first sememes are chosen as classification features. To verify the effectiveness of the proposed method,in the experiment,questions of different size at random are extracted from the corpus to build the incremental sets. Then classify the questions from the same test set based on different incremental sets. Experimental results show that the incremental Bayes classifier achieves better result. The classification accuracy of coarse classes and fine classes achieves 90． 2% and 76． 3% respectively. At the same time, it significantly optimizes the efficiency to some degree.