一种循环神经网络的词义消歧方法  

A Method of Word Sense Disambiguation with Recurrent Netural Networks

作　　者：张春祥 周雪松 高雪瑶[2] ZHANG Chun-xiang;ZHOU Xue-song;GAO Xue-yao(School of Software and Microelectronics,Harbin University of Science and Technology,Harbin 150080,China;School of Computer Science and Technology,Harbin University of Science and Technology,Harbin 150080,China)

机构地区：[1]哈尔滨理工大学软件与微电子学院,哈尔滨150080 [2]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《哈尔滨理工大学学报》2020年第1期80-85,共6页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61502124,60903082);中国博士后科学基金(2014M560249);黑龙江省自然科学基金(F2015041,F201420);黑龙江省普通高校基本科研业务费专项资金(LGYC2018JC014)。

摘　　要：词义消歧是自然语言处理领域中的一个重要研究问题。针对汉语一词多义现象,结合上下文语境,采用循环神经网络(Recurrent Netural Networks,RNN)来确定歧义词汇的真实含义。以目标歧义词汇为中心,提取其左右4个邻接的词汇单元。从中抽取词形、词性和语义类作为消歧特征。以消歧特征为基础,结合循环神经网络来构建词义消歧分类器。利用SemEval-2007:Task#5的训练语料和哈尔滨工业大学语义标注语料来优化RNN的参数。使用SemEval-2007:Task#5的测试语料对词义消歧分类器进行测试。实验结果表明:所提出的方法能够提高词义消歧的准确率。Word sense disambiguation is an important research problem in natural language processing field.For the phenomenon that a Chinese word has many senses,recurrent neural network(RNN)is used to determine true meaning of ambiguous word with its context.Target ambiguous word is viewed as center and its four adjacent word units are extracted.Word,part-of-speech and semantic categories are extracted as disambiguation features.Based on disambiguation features,recurrent neural network is used to construct word sense disambiguation classifier.Training corpus in SemEval-2007:Task#5 and semantic annotation corpus in Harbin Institute of Technology are used to optimize parameters of RNN.Test corpus in SemEval-2007:Task#5 is applied to test word sense disambiguation classifier.Experimental results show that the proposed method can improve accuracy of word sense disambiguation.

关 键 词：词义消歧 自然语言处理 循环神经网络 消歧特征 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]