基于图文关联的威判指标体系自适应构建模型
Adaptive Construction Model of Threat-Judgment Index System Based on Graphic-text Correlation
孙鹏帅 1闫永亮 1杨京雷 2李昊瞳 1周敏敏 1侯帅 1魏永勇3
作者信息
- 1. 河北工程大学信息与电气工程学院,河北邯郸 056038
- 2. 中华通信系统有限责任公司,河北石家庄 050000
- 3. 中国兵器科学研究院,北京 100089
- 折叠
摘要
威判指标体系是评估敌方威胁程度和指导军事决策的重要内容,而威判指标体系的自动构建是一项关键技术.针对目前威判指标体系自动构建面临数据样本少和噪声大的问题,提出了一种基于图文关联的威判指标体系自适应构建模型BiLSTM-CRF-ITAAC.首先,构建了内容关联自适应权重矩阵,对实体模型提取出的三元组进行筛选,得到威判指标体系数据初集;其次,提出了文档威判指标体系图片识别模型,以识别出包含有威判指标体系的图片;最后,提出了威判指标体系图文关联模型,以关联威判指标体系数据初集,生成威判指标体系.经实验验证,所提出的模型与单一实体提取模型相比,可有效降低噪声,这为威判指标体系的构建奠定了重要理论基础.
Abstract
The threat-judgment index system is an important part of evaluating the threat level of the enemy and guiding military decision-making,and the automatic construction of the threat-judgment index system is a key technology. Aiming at the problem of small data samples and high noise in the automatic construction of the threat-judgment index system,an adaptive construction model BiLSTM-CRF-ITAAC based on image-text association was proposed. Firstly,the content association adaptive weight matrix is constructed,and the triples extracted from the solid model are screened to obtain the initial data set of the threat-judgment index system. Secondly,a picture recognition model of the document threat-judgment index system is proposed to identify the pictures containing the threat-judgment index system. Finally,an graphic-text correlation model of the threat-judgment index system is proposed to adjust and correct the initial data set of the threat-judgment index system and generate the threat-judgment index system. The experimental results show that compared with the single entity extraction model,the model proposed can effectively reduce the noise and improve the accuracy,which lays an important theoretical foundation for the construction of the threat-judgment index system.
关键词
威判指标体系/三元组提取/图文关联Key words
threat-judgment index system/triplet extraction/graphic-text correlation引用本文复制引用
基金项目
河北省自然科学基金面上项目(F2021402009)
河北省自然科学基金面上项目(A2020402013)
出版年
2024
NETL
NSTL
万方数据