基于实例的汉语语义超常搭配的自动发现  

Example-based Discovery of Unconventional Chinese Semantic Collocations

作　　者：杨芸[1] 李剑锋[1] 周昌乐[1,2] 黄孝喜[2,3] 

机构地区：[1]厦门大学人工智能研究所,厦门361005 [2]浙江大学语言与认知研究中心,杭州310028 [3]浙江大学计算机科学与技术学院,杭州310027

出　　处：《计算机科学》2008年第9期195-197,共3页Computer Science

基　　金：国家自然科学基金资助(项目号:60373080);浙江大学985工程资助

摘　　要：搭配在语言信息处理中具有重要的应用价值,通常我们主要关注符合语法规则的常规搭配。实际上,语言中还存在着大量的语法上符合规则而语义上不符合常规认知的语义超常搭配现象,而这样的现象与语言的隐喻表达和思维有着密切的联系,对自然语言理解将产生重要的影响。本文面向汉语隐喻理解来研究文本中语义超常搭配的自动发现方法,从汉语语义超常搭配判断的心理机制出发,提出了基于实例的汉语语义超常搭配识别的量化计算方法。实验以动词为中心的搭配语料为测试集,语义超常搭配识别的召回率为80.7%,准确率为81.5%。实验结果表明本文所给出的基于实例语义超常搭配判断的办法是切实可行的。Semantic collocation is the tendency for lexical words to occur together. It is constrained by the syntactic structure and semantic cohesion between its lexical items. Besides conventional semantic collocations, there are also unconventional semantic collocations whose semantic combination is far beyond people＇s conventional cognitive sense. They are pervasive in all types of writing and considered to be the focus of rhetorical and metaphorical expressions which could be a real difficulty in Natural Language Understanding. In this paper,the characteristics of unconventional semantic collocations and their psychological triggering process are first analyzed and an example-based model is then proposed to automatically discover unconventional semantically collocated word pairs. The experiment is carried out on Chinese verbal collocations. The precision is 81.5% and recall is 80. 7%. The experimental results show that the exampie-based model is effective in detecting unconventional semantic collocations.

关 键 词：义超常搭配 语义搭配超常度 隐喻 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP311.13[自动化与计算机技术—计算机科学与技术]