基于语义依存关系的汉语复句关系词自动识别  

Research on Automatic Recognition of Relational Word in Chinese Compound Sentences Based on Semantic Dependencies Analysis

作　　者：杨进才[1] 郑雷 胡泉 YANG Jincai;ZHENG Lei;HU Quan(School of Computer Science,Central China Normal University,Wuhan 430079)

机构地区：[1]华中师范大学计算机学院,武汉430079

出　　处：《计算机与数字工程》2021年第8期1531-1537,1618,共8页Computer & Digital Engineering

基　　金：国家社科基金项目(编号:19BYY092)资助。

摘　　要：复句关系词自动识别是复句层次识别以及对复句语义理解的基础。目前,汉语关系词的自动识别主要根据字面特征与句法特征,未涉及到语义特征的分析。论文利用哈工大的LTP平台,分析句子的语义依存关系,提取语义特征,建立语义依存特征库,运用随机森林方法自动识别复句关系词。复句关系词识别正确率92.32%,F1值为92.31%。实验结果表明了该方法的有效性。The automatic recognition of relational word in Chinese compound sentences is the key to divide hierarchical relations and discriminate the meaning of the whole compound sentence.At present,the automatic recognition of Chinese relational words is mainly based on the literal features and syntactic features,which doesn't involve the analysis of semantic features.In this paper the LTP(Language Technology Platform)developed by harbin institute of technology is used to analyze the semantic dependence of sentences,extract semantic features,establish semantic dependency database,and random forest method is used to automatically identify related words in compound sentence.The correctness of recognition is 92.32%,and the F1 value is 92.31%.Experimental results demonstrate the effectiveness of this method.

关 键 词：复句关系词 语义依存关系 随机森林 自动识别 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]