基于改进编辑距离和依存文法的汉语句子相似度计算  

CHINESE SENTENCE SIMILARITY COMPUTING BASED ON IMPROVED EDIT-DISTANCE AND DEPENDENCY GRAMMAR

作　　者：刘宝艳[1] 林鸿飞[1] 赵晶[1] 

机构地区：[1]大连理工大学计算机科学与工程系,辽宁大连116024

出　　处：《计算机应用与软件》2008年第7期33-34,47,共3页Computer Applications and Software

基　　金：国家自然科学基金资助项目(60373095)

摘　　要：句子相似度计算在中文自然语言处理领域有着广泛的应用背景。要准确地刻画一个句子所表达的意思,必须深入到语义层面级并结合语法结构信息,提出了一种基于改进编辑距离和依存文法的汉语句子相似度计算方法。依存文法考虑到句子内部的结构和词语之间的相互作用关系,而编辑距离由于《同义词词林》的应用可以兼顾同义词之间的替换,因此该方法与其他方法相比,描述句子的信息更加全面,试验结果表明该方法是有效的。Sentence similarity computing has wide application background in the field of Chinese natural language processing. For describing accurately the meaning of a sentence, the deep study must be done at semantic level as well as considering its features of grammatical structure. The paper proposes an approach for computing sentence similarity based on Improved Edit-Distance and Dependency Grammar. Dependency Grammar refers to the structure inside a sentence and the relations among phrases and words, and Edit-distance can take account of the substitution of synonyms based on a dictionary ＂Synonymy Thesaurus＂. Comparing with other methods, this method fully describes the features of the sentence. The experiments also showed that it improved the accuracy percentage.

关 键 词：相似度计算 依存文法 改进编辑距离 句法分析 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP391[自动化与计算机技术—计算机科学与技术]