山西大学学报（自然科学版）2025，Vol.48Issue(1)：20-28,9.DOI:10.13451/j.sxu.ns.2024133
基于中心事件和句子感知的事件因果识别模型
Event Causal Identification Model Based on Central Event and Sentence Perception
摘要
Abstract
In the financial sector,recognizing causal relationships between events at the discourse level is a challenging task.Event causality is linked to central events and their corresponding sentences.On the one hand,central events among multiple events influ-ence the causal relationships among them;on the other hand,the sentences in which events occur can also reinforce these causal links.To address this issue,this paper proposes a central event and sentence-aware model for event causality recognition.The model uses the Centrality-aware High-order Event Reasoning Network(CHEER)to establish the representation of the central event,and constructs the event relationship graph based on the graph attention network.This graph captures the representations of events and event pairs,adaptively modeling causal transmission among events,thereby identifying central events and their causally related counterparts.By introducing sentiment-centered and cue word-centered embeddings,the model enhances the representation of sen-tences'sentiment and structure,which improves the accuracy of causal relationship recognition in financial texts.Experiments con-ducted on the publicly available EventStoryLine dataset,annotated with central events,and a self-constructed dataset showed that the proposed model improved inter-sentence accuracy by 1.1％in the EventStoryLine dataset and achieved an overall improvement of 1.8％in the self-constructed dataset.These results confirm that central events significantly enhance the accuracy of event causali-ty recognition.关键词
事件因果关系/事件关系图/篇章级文本/金融领域/图注意力网络Key words
event causality/event relationship graph/document-level text/financial domain/graph attention network分类
信息技术与安全科学引用本文复制引用
尚舒鹏,普瑞丽,李旸,王素格,李德玉..基于中心事件和句子感知的事件因果识别模型[J].山西大学学报（自然科学版）,2025,48(1):20-28,9.基金项目
国家自然科学基金(62106130 （）
62072294 （）
62376143) （）
山西省青年科学研究项目(20210302124084) （20210302124084）
山西省高等学校科技创新项目(2021L284) （2021L284）
