摘要
情报数据存在多源异构、关联缺失、重复冗余等问题,有限的数据处理能力已经无法满足不断增长的数据获取能力。事件共指消解任务旨在将互为共指关系的事件识别为同一事件并进行融合处理。对融合多源情报进行研究,提出一种端到端的事件共指消解方法。从情报文本中自动抽取情报事件;编码整个情报文档得到待消解事件的表示,计算每对事件提及的共指得分,以此构建文档内事件共指链;通过算法利用文档内事件共指链融合多源情报文档中的共指事件。实验结果表明,提出方法对消除冗余信息、简化情报文本、融合情报信息具有明显增益。
Intelligence data present problems such as multi-source heterogeneity,lack of information association,and a large number of redundant events.Limited data processing ability can no longer meet the increasing data acquisition ability.Event coreference resolution is to aggregate the coreferential events into the event chain.An end-to-end event co-reference resolution method was proposed based on multi-source information fusion.Intelligence events were automatically extracted from intelligence texts.The whole intelligence document was encoded to obtain the representation of the events to be resolved,and the coreference score of each pair of events was calculated to construct the coreference chain of the events in the document.The result of within-document coreference resolution was utilized to integrate the coreferential events in multi-source intelligence documents.Experimental results show that the proposed method is of great help to eliminate redundant information,simplify information text and fuse information.
作者
环志刚
蒋国权
周泽云
陈涛
严浩
HUAN Zhi-gang;JIANG Guo-quan;ZHOU Ze-yun;CHEN Tao;YAN Hao(The Sixty-Third Research Institute,National University of Defense Technology,Nanjing 210007,China;School of Cyber Science and Engineering,Southeast University,Nanjing 210096,China;Information Center of Equipment Development Department,Beijing 100034,China;Equipment Development Department,Beijing 100034,China)
出处
《计算机工程与设计》
北大核心
2023年第10期3124-3131,共8页
Computer Engineering and Design
关键词
多源情报
信息融合
端到端
事件共指消解
文档内
跨文档
神经网络
multi-source intelligence
information fusion
end-to-end
event coreference resolution
within-document
cross-document
neural network