基于图注意力网络的科技实体关系联合抽取方法  

Joint Extraction of Science Entity and Relation Based on Graph Attention Network

作　　者：周义恒 陈颢天 周向东[1,2] ZHOU Yi-heng;CHEN Hao-tian;ZHOU Xiang-dong(College of Computer Science and Technology,Fudan University,Shanghai 200438,China;Shanghai Key Laboratory of Data Science,Shanghai 200438,China)

机构地区：[1]复旦大学计算机科学与技术学院,上海200438 [2]上海市数据科学重点实验室,上海200438

出　　处：《工程管理科技前沿》2022年第3期39-44,共6页Frontiers of Science and Technology of Engineering Management

基　　金：国家重点研发计划资助项目(2018YFB1402600)。

摘　　要：实体关系抽取在科技文献挖掘、文本知识发现等应用中起着重要作用。本文提出了一种新的实体关系联合抽取模型,利用关系之间存在的关联性来提升实体与关系联合抽取的性能。即将实体二元组(实体对)映射到关系语义子空间,并利用图注意力网络(Graph Attention Network)来探索关系之间相关性,实现了实体关系三元组抽取性能的改进。在两个常用的公开数据集上的实验结果表明本模型达到了目前已知的最好性能指标。Entity and relation extraction plays an important role in science text mining and science knowledge and information discovery.In this paper,we propose a novel entity and relation joint extraction model,utilizing the association information between relations to improve model performance.Specifically,we project entity pair vectors to corresponding relation-spaces,and adopt Graph Attention Network(GAT)to adaptively explore the association information between them,thereby improving the model performance.The experimental results on two commonly used datasets show that our proposed model outperforms other existing models and archives the state-of-the-art performance.

关 键 词：关系相关性 联合抽取 图注意力网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]