一种汉语语句依存关系网协动生成方法研究  

A Study of the Dynamic Generative Methods For Dependency Relation Network of Chinese Sentences

作　　者：郭艳华[1] 周昌乐[2] 

机构地区：[1]杭州电子工业学院信息工程学院,浙江杭州310037 [2]浙江大学计算机系,浙江杭州310027

出　　处：《杭州电子工业学院学报》2000年第4期24-32,共9页Journal of Hangzhou Institute of Electronic Engineering

基　　金：国家自然科学基金的资助!(6 99830 0 6 )

摘　　要：句法分析的研究是自然语言理解领域的一个重要组成部分。但是汉语在词类这个语言层次上存在着许多歧义结构 ,这给汉语的自动句法分析带来了难以逾越的障碍。本文以依存语法作为语言模型的基础 ,提出了一种全新的句子分析策略 :分析在两级上进行 ,一级是基于中心动词同其他成分间的约束关系 ,通过寻找汉语语义类之间可能存在的句法关系 ,实现句子成分过滤 ,完成句子主干提取。一级是上下文级 ,将语法、语义和语境信息一体化 ,结合依存语法确定汉语句子中各成分间的依存关系。通过实验表明了该分析策略的可行性。The research of parsing is important in the field of natural language processing.But there are too many syntactic ambiguities in Chinese,and it is not easy to parse the Chinese sentences automatically only with the Part of Speech tagging.In this paper,taking Dependency Grammar as language processing model, we put forward a new strategy of Chinese sentence components analysis. Strategy of parsing works on sentence level and context level.On sentence level,a sentence consititute filtering algorithem has been implemented to extract sentence frame. On context level,complatible use of syntactic and semantic information,to obtain the dependency relations between different components based on Dependency Grammar. The result of experiment shows that this strategy is feasible not only in theory but also in practice.

关 键 词：句法分析 依存语法 歧义 语法 语义 语境 

分 类 号：TP301.4[自动化与计算机技术—计算机系统结构]