词义刻画与词义识别  

Discrimination of semantic depiction and statistical word sense

作　　者：范冬梅[1] 卢志茂 张汝波 

机构地区：[1]哈尔滨工程大学模式识别与自然计算研究室,黑龙江哈尔滨150001

出　　处：《计算机工程与设计》2011年第1期270-273,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(60603092;60975042);国家教育部博士点专向基金项目(20070217043)

摘　　要：为了改善词义识别的效果,使词义识别技术水平获得实质性的提高,需要考察词义刻画与词义理解之间的关系,确定是否有必要重新刻画词语的词义,需要从理论和实验角度进行探讨。从语言学角度分析了影响词义刻画的各种因素,并设计和实现了人机词义识别的对比实验,借以揭示词义刻画粒度大小与词义识别精度之间的关系,进而说明调整词义刻画粒度的必要性。实验结果表明,科学控制词义刻画的粒度可以增强词义的可计算性,提高词义识别的精度。In order to improve the effect of word sense discrimination,to obtain substantial improvement,we need to look at the relationship between semantic characterization and word sense understanding,need to research these problems by the theore and experiment. This article analyzes the impact of various factors that characterize word sense in the view of Linguistics.Furthermore,we design and implement the comparative experiments between human and machine word sense discrimination.Those experiments reveal the relationship between granularity of semantic depiction and accuracy of word sense discrimination.Then it shows the necessity of adjusting the size of semantic characterization.Experimental results show that scientifically controls the granularity of semantic depiction can enhance the computability of word senses,and improve the accuracy of word sense discrimination.

关 键 词：词义刻画 词义识别 计算语言学 自然语言理解 人工智能 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]