软件导刊2026，Vol.25Issue(3)：69-77,9.DOI:10.11907/rjdk.251063
基于桥实体增强的联合事件关系抽取
Joint Event Relation Extraction Based on Bridging Entities Enhancement
摘要
Abstract
Event Relation Extraction(ERE)is a critical task in the field of Natural Language Processing(NLP),aiming to identify semantic relationships between event pairs from text,with broad applications in event reasoning,knowledge graph construction,and information extrac-tion.However,existing methods predominantly rely on event triggers for relation modeling,struggling to accurately capture implicit event rela-tions and complex cross-sentence contextual information.Moreover,the potential value of event arguments,such as bridging entities,has not been fully explored,limiting comprehensive modeling of event relations.To address these challenges,this paper proposes a bridging entity-en-hanced joint event relation extraction method.This approach leverages NLP tools to extract intra-sentence and cross-sentence entities of event triggers as bridging entities,jointly modeling them with cross-attention and dual gating mechanisms.The cross-attention module captures global interaction information between triggers and bridging entities,while the dual gating mechanisms dynamically fuse the interaction infor-mation between event triggers and cross-attention outputs,as well as adjust the fusion of two event representations.The method was evaluated on the MAVEN-ERE dataset through experiments and comparisons,demonstrating superior performance over baseline models on causal,tem-poral,subevent,and coreference relations,thus validating its effectiveness.关键词
事件关系抽取/桥实体/交叉注意力机制/桥实体抽取算法/自然语言处理Key words
event relation extraction/bridging entities/cross-attention mechanism/bridging entities extraction algorithm/natural language processing分类
信息技术与安全科学引用本文复制引用
张俊驰,刘浩源,冼彦宁,朱珣..基于桥实体增强的联合事件关系抽取[J].软件导刊,2026,25(3):69-77,9.基金项目
国家自然科学基金青年科学基金项目(62106179) （62106179）
