基于E-Chunk的问句实例分析系统  

Question Instance Analysis System Based on E-Chunk

作　　者：骆正华[1] 樊孝忠[1] 刘林[1] 龚永罡[1] 

机构地区：[1]北京理工大学信息科学技术学院计算机科学工程系,北京100081

出　　处：《北京理工大学学报》2005年第1期63-66,共4页Transactions of Beijing Institute of Technology

摘　　要：分析中文问句的结构特点,研究处理问句需要解决的问题,提出一种基于语义块实例分析问句的新方法.采用语义块向量表示问句实例,比较输入问句和问句实例库中问句实例的相似性.在语义块向量结构相似的基础上计算问句的相似度.测试结果表明,该方法可行,准确率和召回率可分别达到82.05%和91.95%,该方法对问句分析系统的设计具有借鉴意义和继续深入研究的价值.The structure characteristic of the Chinese question sentence is analyzed, the problem that needs to be solved when analyzing the question sentence is studied and dealt with. A kind of new method based on question sentence instance is proposed. The E-Chunk vector to express the instance of the question sentence is adopted. The similarity between the input sentence and the sentence in storehouse is discussed. The degree of similarity of the question sentence is calculated on the basis of the similarity of the semantic chunk vector structure. The experiment indicates that the method is feasible, the rate of accuracy and recalling rate can be up to 82.05% and 91.95% respectively. The experiments prove that it is feasible to use the method to analyze question clause and it is valuable to continue to study the method deeply.

关 键 词：语义块 问句实例 语义相似度 知网 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]