面向数据分析的中文问句句法研究  

Research of the Chinese Question Syntax Based on Data-Oriented Parsing

作　　者：郭海旭[1] 吕国英[1] 李茹[1] 

机构地区：[1]山西大学计算机与信息技术学院,山西太原030006

出　　处：《计算机工程与科学》2009年第10期101-104,共4页Computer Engineering & Science

基　　金：国家863计划资助项目(2006AA01Z142);国家社会科学基金青年项目(07CYY022)

摘　　要：本文以面向数据分析为基本思想,同时结合相似度概率评估技术,实现中文问句句法分析。对于输入问句,首先需要经过词汇层与词性层以及句型和疑问词初选得到片段;然后,基于已构建的数据库,对词汇词性序列的子序列搜索片段,进行片段组合;最后,对输入语句与初选结果进行相似性评估,完成输入语句的组合分析过程。为论证方法有效性,基于200句山西旅游问句做实验。实验结果表明,句法分析的准确率达到96%。The paper makes use of the data-oriented parsing technology and the similarity-based probability estimation technique to achieve the Chinese question syntax parsing. At first, every input sentence must be primarity selected, then it executes the search for the databases based on the word level and the part-of-speech level. Finally, the similarity estimation between the input sentence and the initial selection result is performed using the similarity-based probability estimation technique. So the combination parsing process of the input sentence can be completed successfully. To prove the efficiency of the method,the database source is constructed based on the 200 Shanxi tourism questions. The experimental results show that the accuracy rate is 96%.

关 键 词：面向数据的分析 句法分析 相似性评估 句法截断 句法片段 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]