英语花园幽径句的线性剖析:基于斯坦福解析器的分析  

Towards Linear Parsing of English Garden Path Sentences:An Analysis Based on Stanford Parser

作　　者：杜家利 于屏方[3] 

机构地区：[1]广东外语外贸大学语言工程与计算实验室,南京210093 [2]南京大学外国语学院,南京210093 [3]广东外语外贸大学中国语言文化学院,广州510420

出　　处：《鲁东大学学报（哲学社会科学版）》2017年第5期39-46,共8页Ludong University Journal:Philosophy and Social Sciences Edition

基　　金：国家留学基金项目(201507150009)&(201607150007);中国博士后特别资助项目(2016T90440)和一等资助项目(2015M570424);广东省社科项目(GD15HWW02)

摘　　要：本文讨论了美国斯坦福解析器在英语花园幽径句线性剖析中的不足。斯坦福解析器在句法上采用词汇化的概率上下文无关语法,在语义上使用词汇化的依存分析,其概率策略的使用极大地提高了系统剖析的效率和能力。但是,这种概率分析器在处理具有局部歧义效应的英语花园幽径句时却出现了系统性错位。本文从词类标注、句法剖析和依存关系分析入手,对低概率的英语花园幽径句的系统剖析错位进行了较全面的分析。结论是:斯坦福解析器对低概率英语花园幽径句的线性剖析不敏感,容易受句式的线性发展关系和认知模式的影响,其解码效率尚有提高的余地。This article discusses the shortage of Stanford Parser in the linear processing of English garden path sentences. Stanford Parser uses both lexicalized probabilistic context free grammar in syntax and lexicalized dependency in semantics. Its probabilistic strategy increases the efficiency and capacity in the systematic parsing greatly. However,this probability-based analyzer may experience the processing breakdown when dealing with English garden path sentences with local ambiguity effect. From the perspective of part of speech tagging,syntactic parsing and dependency analysis,this article comprehensively analyzes the processing breakdown resulting from English garden path sentences with low probability. It draws a conclusion that Stanford Parser is not sensitive to linear parsing of English garden path sentences with low probability,the parsing result is severely affected by syntactic pattern＇s linear development and cognitive models,and there is still room for improvement in Stanford Parser＇s systematic decoding.

关 键 词：计算语言学 花园幽径模式 行进错位 局部歧义 斯坦福解析器 

分 类 号：H087[语言文字—语言学]