篇章级文本的事件图谱构建与事件抽取方法  

Document-level event graph construction and event extraction method for text data

作　　者：张源奔 赵一斐 赵志浩 韩众和 张利利 谢辉平[1,2] ZHANG Yuan-ben;ZHAO Yi-fei;ZHAO Zhi-hao;HAN Zhong-he;ZHANG Li-li;XIE Hui-ping(The Aerospace Information Research Institute,Chinese Academic of Sciences,Beijing 100190,China;Key Laboratory of Network Information System Technology(NIST),Chinese Academy of Sciences,Beijing 100190,China;Aerospace Information Research Institute of QiLu,Chinese Academic of Sciences,Jinan 250000,China;College of Information Science and Technology,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]中国科学院空天信息创新研究院,北京100190 [2]中国科学院网络信息体系技术重点实验室,北京100190 [3]中国科学院齐鲁空天信息研究院,山东济南250000 [4]北京化工大学信息科学与技术学院,北京100029

出　　处：《计算机工程与设计》2024年第1期146-152,共7页Computer Engineering and Design

摘　　要：针对篇章级事件抽取任务主要存在的以下3个挑战:模型复杂、事件论元分散以及多事件抽取,提出一种基于异质交互图和有序树的篇章级事件抽取方法DTHIGN。支持以解耦实体抽取部分的方式精简模型的参数;构建面向触发词的篇章级异质交互图,更全面地对语义进行建模;支持使用基于统一事件模板的树形事件论元抽取方式获得事件论元结果。实验结果表明,该方法能够获得较好的效果并显著降低模型的训练难度。To solve three main challenges in the task of document-level event extraction(DEE),namely model complexity,scattered event arguments,and multi-events extraction,a DEE method DTHIGN based on heterogeneous interaction graphs and order tree was proposed.This method supported reducing parameters of itself by decoupling entity extraction part.Document-level trigger-oriented heterogeneous interaction graph was constructed to model semantics more comprehensively.The tree-based event argument extraction method with unified event template was used to obtain event arguments results.Experimental results show that the proposed method achieves better performance and significantly reduces the training complexity.

关 键 词：深度学习 自然语言处理 事件图谱 事件抽取 事件论元抽取 异质图 图卷积网络 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]