金融领域事件因果关系发现及事理图谱构建与应用  

Event Causality Extraction, Eventic Graph Construction and Application in Financial Domain

作　　者：杨纪星 杨波[1,2] 朱剑林 康怡琳[1,2] YANG Jixing;YANG Bo;ZHU Jianlin;KANG Yilin(College of Computer Science,South-Central Minzu University,Wuhan,Hubei 430074,China;Key Laboratory of Cyber-Physical Fusion Intelligent Computing(South-Central Minzu University),State Ethnic Affairs Commission,Wuhan,Hubei 430074,China)

机构地区：[1]中南民族大学计算机科学学院,湖北武汉430074 [2]信息物理融合智能计算国家民委重点实验室,湖北武汉430074

出　　处：《中文信息学报》2023年第7期131-142,共12页Journal of Chinese Information Processing

基　　金：国家自然科学基金(72104254);国家重点研究与发展计划(2020YFC1522600);湖北省自然科学基金(2022CFB469)。

摘　　要：事理图谱是研究事物动态发展的有效手段。针对金融因果事理图谱构建过程中数据集匮乏及构建方案缺少实践对比的现状,该文面向金融领域中发生频率较高的热点事件,研究构建事理图谱的方法。该文提出了一种新的金融领域事件论元的定义,制定了基于ATT+SBV结构的句法分析方案,针对信息抽取任务提出了面向金融因果事件的序列标注定义。该文同时提出了一种基于BERT+Bi-LSTM+CRF模型的信息抽取方案,并与不同神经网络模型进行对比研究。实验结果表明,该模型在信息抽取任务中,F_1值达到95.78%,准确性有较大提升。该文通过Neo4j图数据库存储并构建金融因果事理图谱,以事件关系可视化的方式揭示现实金融事件的演变逻辑规律,分析金融网络的风险传导扩散机制。The eventic graph is an effective solution to study the dynamic development of things.Aiming at the construction of financial causal eventic graph,this paper explores the methods of constructing eventic graph for the headlines with high frequency in the financial field.In this paper,we propose a new definition of event argument in the financial domain,a syntactic analysis scheme based on ATT+SBV structure,and a definition sequence labeling for financial causal events based on information extraction task.We also propose an information extraction scheme based on BERT+Bi-LSTM+CRF model,and compares it with different neural network models.The experimental results show that the proposed method achieves a significant improvement of F1 value of 95.78%.With Neo4j graph database,we demonstrate the event relationship visualization to reveal the evolution of real financial events,as well as the risk transmission and diffusion mechanism of financial network.

关 键 词：事理图谱 事件抽取 信息抽取模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]