基于滑动语义串匹配(SMOSS)的汉语词义消歧  

Chinese Word Sense Disambiguation Based on Sliding Match of Semantic String(SMOSS)

作　　者：王伟[1] 黄德根[1] WANG Wei;HUANG De-gen(School of Computer Science and Technology,Dalian University of Technology,Dalian 116033,China)

机构地区：[1]大连理工大学电信学部计算机科学与技术学院,辽宁大连116033

出　　处：《小型微型计算机系统》2020年第7期1345-1350,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61672127,U1936109)资助。

摘　　要：为了提高汉语词义消歧的性能,提出了一种基于滑动语义串匹配(Sliding Match of Semantic String,SMOSS)的汉语词义消歧方法.首先,从标注词义的训练语料中提取N元语义模板,建立N元语义模板库;之后,从待消歧句子中提取N元语义码串与N元语义模板库中的语义模板匹配,通过计算匹配成功的多个模板的得分来确定歧义词的最终词义.该方法具有好的弹性匹配能力和宽的匹配范围,能够有效减少数据稀疏问题.实验采用了SemEval2007-Task#5中文词义消歧的评测标准,消歧正确率为75.06%,与目前已知的最好系统性能相近.In order to improve the performance of Chinese word sense disambiguation(WSD),this paper proposes a method for Chinese word semantic disambiguation based on sliding match of semantic string(SMOSS).Firstly,some N-gram semantic templates are extracted from corpus annotated by semantic code and then one N-gram semantic template bank is built.Secondly,each N-gram semantic code string is extracted from test sentence to match with the N-gram templates in the N-gram semantic template bank,and one final semantic code is decided by choosing the one with highest score that combines the votes of multiple matched templates.This method has better matching elasticity and wider matching range than traditional word template matching,reduces the problem of data sparsity.In the experiment,the evaluation standard of SemEval2007-Task#5 is adopted,and the accuracy is 75.06%,which is close to the best know n system performance.

关 键 词：词义消歧 N元语义模板 滑动语义串匹配 SMOSS 

分 类 号：TP305[自动化与计算机技术—计算机系统结构]