快速检索:      
引用本文:
【打印本页】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
过刊浏览    高级检索
本文已被:浏览 99次   下载   
分享到: 微信 更多
问答系统中基于语义核函数的问题分类算法
江龙泉,张波,胡志鹏,丁峻宏,刘波
作者单位E-mail
江龙泉 上海师范大学 chris.lq.jiang@gmail.com 
张波 上海师范大学 zhangbo@shnu.edu.cn 
胡志鹏 上海师范大学  
丁峻宏 上海超算科技有限公司  
刘波 上海超算科技有限公司  
摘要:
问题分类是问答系统中的一个重要组成部分。传统的基于支持向量机的问题分类算法采用问题的Bag-of-Words(BOW)特征,虽然简单有效,但当任务需要更多的语义信息时则无法对问题特点进行更加深入的分析,从而影响分类效果。提出一种基于语义核函数的问题分类算法,该算法基于问题的语法结构构建SVM核函数。首先,将给定的问题解析为语法树结构,用语法树的子树表示该问题;然后,从词法、语法、语义三个层面提取问题的特征,构成更加丰富的特征空间;接着,基于问题的语法树构建核函数;最后,使用潜在语义索引方法并结合问题的词法、语法以及语义特征,通过语义核函数将特征空间映射到更有效的空间中进行问题分类。TREC数据集上的实验结果表明,通过词法、语法以及语义增强的问题特征空间可以提高分类准确率。
关键词:  问答系统  监督学习  SVM  问题分类  语义核函数  特征空间
DOI:
分类号:
基金项目:国家自然科学基金项目(面上项目,重点项目,重大项目)
A Semantic Kernel Function Based Question Classification Algorithm for Question Answering System
JIANG Longquan,ZHANG Bo,HU Zhipeng,DING Junhong,LIU Bo
Abstract:
Abstract: Question classification is an important part of the Q&A system. Traditional algorithms based on SVMs using the bag-of-words(BOW) features are effective and simple, but in the task in which many aspects of features including semantics are required, they have a significant weakness of being unable to present a deep analysis on a question. An algorithm based on semantic kernel functions for question classification is proposed, and the kernel is constructed using the syntactical structure of the question. Firstly, a given question is parsed into its syntactical structural tree, and then sub-trees of the syntactical tree are used to represent the question. Secondly, features are extracted from three aspects of the question: lexical, syntactical and semantic, and are formed into richer feature space. Thirdly, the kernel is constructed using the syntactical structural tree of the question. Finally, using the potential semantic indexing method, the feature space is mapped into a more efficient space by the potential semantic kernel. The experimental results on the TREC dataset show that the classification accuracy can be improved by lexical, grammatical, and semantic enhancement.
Key words:  Question Answering  Supervised Learning  SVM  Question Classification  Semantic Kernel Function  Feature Space