恐怖袭击事件命名实体识别研究  

Research on Named Entity Recognition of Global Terrorism Events

作　　者：李林瑛[1] 王孙和 卢睿 陈志浩 LI Linying;WANG Sunhe;LU Rui;CHEN Zhihao(School of Software Engineering,Dalian University of Foreign Languages,Dalian Liaoning 116044,China;Public Security Information Department,Liaoning Police College,Dalian Liaoning 116036,China;Asiainfo Security Technologies Co.,Ltd.,Nanjing Jiangsu 210012,China)

机构地区：[1]大连外国语大学软件学院,辽宁大连116044 [2]辽宁警察学院公安信息系,辽宁大连116036 [3]亚信安全科技股份有限公司,江苏南京210012

出　　处：《辽宁警察学院学报》2025年第2期91-96,共6页Journal of Liaoning Police College

基　　金：2022年辽宁省研究生教育教学改革研究项目“《网络数据采集与处理》实践教学模式创新研究”(LNYJG2022423);辽宁省教育厅高等学校基本科研项目“基于深度学习的涉恐事件抽取模型研究”(LJKMZ20221549)。

摘　　要：恐怖袭击事件文本中含有丰富的涉恐实体,如涉恐人物、恐怖袭击目标、恐怖袭击组织等。恐怖袭击文本中实体往往较为复杂,传统的静态词向量的命名实体识别模型存在分词误差传递问题,并且无法表征不同语境中存在的一词多义。针对这些问题,本文提出结合了BERT及全局指针网络的涉恐文本命名实体识别模型,通过BERT学习结合上下文的动态词向量表示,再由双向长短时记忆网络获取文本信息,采用全局指针识别备选涉恐实体片段,并送入实体分类器得到最终实体类别。实验结果表明:该模型的F1值达到92.01%,在涉恐语料的命名实体识别任务中具有有效性。Texts related to terrorist attacks contain rich information about entities involved,such as individuals,targets,and organizations.The identification of entities in terrorist attack texts tends to be rather complex.Traditional named entity recognition models using static word vectors often suffer from segmentation errors and struggle to address the issue of polysemy in different contexts.This paper proposes a named entity recognition model for terrorist-related texts that combines BERT with a global pointer network.By leveraging BERT to learn dynamic word vector representations within context and utilizing bidirectional long short-term memory networks to capture textual information,the model identifies candidate terrorist entity segments through global pointers and subsequently assigns them final entity categories via an entity classifier.Experimental results demonstrate that the model achieves an F1 score of 92.01%,validating the effectiveness of this model in addressing the task of named entity recognition in terrorist attack event corpora.

关 键 词：恐怖袭击 命名实体识别 全球恐怖主义数据库 全局指针 

分 类 号：D918.2[政治法律—法学]