基于同义词词林的句子语义相似度方法及其在问答系统中的应用  

A SEMANTIC SIMILARITY METHOD OF SENTENCES BASED ON TONGYICI CILIN AND ITS APPLICATION IN QUESTION ANSWERING SYSTEM

作　　者：周艳平[1] 李金鹏 蔡素 Zhou Yanping;Li Jinpeng;Cai Su(College of Information Science and Technology,Qingdao University of Science and Technology,Qingdao 266061,Shandong,China)

机构地区：[1]青岛科技大学信息科学技术学院

出　　处：《计算机应用与软件》2019年第8期65-68,81,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(61402246);山东省高等学校科技计划项目(J14LN31)

摘　　要：提出一种基于同义词词林的句子语义相似度方法,借助同义词词林来计算句子的词形相似度,使用向量距离法得到句子间的词序相似度。同时,对句子进行语义依存句法分析。通过对词形、词序、语义依存相似度加权结合获得句子之间的最终相似度。将该方法应用于常问问题问答系统(Frequency Asked Questions,FAQ)的问句匹配。实验结果表明,该方法在问句匹配上相比传统方法具有更高的准确率。This paper proposed a method of sentence semantic similarity based on TongYiCi CiLin,which used TongYiCi CiLin to calculate the morphological similarity of sentences.The similarity of word order between sentences was obtained by using vector distance method.Meanwhile,the sentences were parsed with semantic dependency.The final similarity between sentences was obtained by weighting the similarity of morphological,word order and semantic dependencies.The proposed method was applied to the question matching of Frequency Asked Questions.Experimental results show that this method has higher accuracy in question matching than traditional methods.

关 键 词：问答系统 句子相似度 同义词词林 语义依存 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]