首页 | 本学科首页   官方微博 | 高级检索  
     检索      

面向句法块向量的句子相似度计算方法
引用本文:高顺峰,张再跃.面向句法块向量的句子相似度计算方法[J].教育技术导刊,2009,19(10):106-110.
作者姓名:高顺峰  张再跃
作者单位:江苏科技大学 计算机学院,江苏 镇江 212003
基金项目:国家自然科学基金项目(61371114)
摘    要:传统句子相似度算法没有全面考虑句子结构与语义特征,影响相似度计算准确性,对此提出一种基于句法块向量的句子相似度计算方法。该方法综合考虑句子的语义信息与结构信息,首先构建两句子的语义依存关系树,然后进行一些被动转换等操作,最后根据词向量构建各个句法块向量并通过余弦值计算句子相似度。在常规句子对中进行测试实验,结果表明,综合句子结构与语义信息可提高相似度计算准确性。一般句子相似度计算正确率达到92%,比传统方法提高8%~10%。

关 键 词:句子相似度  语义依存树  词向量  自然语言处理  句法结构  
收稿时间:2020-01-09

Sentence Similarity Calculation Method Based on Syntax Block Vector
GAO Shun-feng,ZHANG Zai-yue.Sentence Similarity Calculation Method Based on Syntax Block Vector[J].Introduction of Educational Technology,2009,19(10):106-110.
Authors:GAO Shun-feng  ZHANG Zai-yue
Institution:School of Computer, Jiangsu University of Science and Technology, Zhenjiang 212003,China
Abstract:Traditional sentence similarity algorithms do not fully consider the structure and semantic characteristics of sentences, which affects the accuracy of similarity calculation. In this regard, a new calculation method for sentence similarity based on syntactic block vectors is proposed. The feature of this method is to comprehensively consider the semantic and structural information of the sentence. It first constructs the semantic dependency tree of the two sentences, then performs some important operations, such as passive conversion, etc., and finally constructs each syntactic block vector and sentence vector based on the word vector sentence similarity is calculated from the cosine value. Tested in regular sentence pairs, the experimental results show that the comprehensiveness of sentence structure and semantic information can improve the accuracy of similarity calculations. For general sentence similarity calculations, the accuracy rate reaches 92%, which is 8% to 10% higher than the traditional method.
Keywords:sentence similarity  dependency syntax tree  word embedding  natural language processing  syntactic structure  
点击此处可从《教育技术导刊》浏览原始摘要信息
点击此处可从《教育技术导刊》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号