中文连动句语义关系识别研究  

Semantic Relation Recognition of Chinese Serial-verb Sentences

作　　者：孙超 曲维光[1,2,4] 魏庭新 顾彦慧[2] 李斌[4] 周俊生[2] SUN Chao;QU Weiguang;WEI Tingxin;GU Yanhui;Li Bin;ZHOU Junsheng(Zhongbei College,Nanjing Normal University,Danyang,Jiangsu 212334,China;School of Computer and Electronic Information/School of Artificial Intelligence,Nanjing Normal University,Nanjing,Jiangsu 210023,China;International College for Chinese Studies,Nanjing Normal University,Nanjing,Jiangsu 210097,China;School of Chinese Language and Literature,Nanjing Normal University,Nanjing,Jiangsu 210097,China;Kunshan Zhenchuan Senior High School,Suzhou,Jiangsu 215300,China)

机构地区：[1]南京师范大学中北学院,江苏丹阳212334 [2]南京师范大学计算机与电子信息学院/人工智能学院,江苏南京210023 [3]南京师范大学国际文化教育学院,江苏南京210097 [4]南京师范大学文学院,江苏南京210097 [5]昆山震川高级中学,江苏苏州215300

出　　处：《中文信息学报》2024年第5期22-31,共10页Journal of Chinese Information Processing

基　　金：国家社会科学基金(21&ZD288)。

摘　　要：连动句是形如“NP+VP1+VP2”的句子,句中含有两个或两个以上的动词(或动词结构)且动词的施事为同一对象。相同结构的连动句可以表示多种不同的语义关系。该文基于前人对连动句中VP1和VP2之间的语义关系分类,标注了连动句语义关系数据集,基于神经网络完成了对连动句语义关系的识别。该方法将连动句语义识别任务进行分解,基于BERT进行编码,利用BiLSTM-CRF先识别出连动句中连动词(VP)及其主语(NP),再基于融合连动词信息的编码,利用BiLSTM-Attention对连动词进行关系判别,实验结果验证了该文所提方法的有效性。Serial-verb sentence contains two or more verbs(or verb structures)sharing the same agent with the form of“NP+VP1+VP2”.Sentences with the same serial-verb structures can express a variety of different semantic relation.In this paper,we build a data set of serial verb sentences,and propose a neural network model for recognition of semantic relation in serial verbs.First,the serial-verb sentences are encoded with Bert.Second,BiLSTM-CRF is used to identify the serial verbs and their subjects.Then,based on the embedding fused with serial verbs,the relation recognition of serial verbs is implemented using BiLSTM-Attention.The experimental results prove the effectiveness of the proposed method.

关 键 词：连动结构 神经网络 连动句语义关系识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]