基于增量式贝叶斯模型的中文问句分类研究  

Chinese Question Classification Research Based on Incremental Bayes Model

作　　者：王小林[1] 镇丽华 杨思春[1] 邰伟鹏[1] 郑啸[1] 

机构地区：[1]安徽工业大学计算机科学与技术学院,安徽马鞍山243002

出　　处：《计算机工程》2014年第9期238-242,共5页Computer Engineering

基　　金：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)

摘　　要：固定训练集生成的分类器性能不理想且不能跟踪用户需求,为此,提出一种将增量式贝叶斯思想用于问句分类的方法。采用遗传算法选取最优特征子集优化分类器,从而避免训练集特征过分冗余,使分类器在学习过程中动态地扩大训练集并修改分类器参数。在对问句进行分类时,提取问句的疑问词、句法结构、疑问意向词和疑问意向词在知网的首项义原作为分类特征。为了验证增量式贝叶斯方法的有效性,从语料库中随机抽取不同规模的问句构成增量集,基于不同的增量集对同一测试集中的问句进行分类。实验结果表明,增量式贝叶斯分类器较朴素贝叶斯分类器有更高的分类精度,大类和小类的准确率分别达到90.2%和76.3%,在提高准确率的同时优化了运行效率。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 achieves90.2% and76.3% respectively. At the same time,it significantly optimizes the efficiency to some degree.

关 键 词：问句分类 问答系统 增量式贝叶斯 朴素贝叶斯 改进贝叶斯 遗传算法 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]