Modern Defense Technology ›› 2024, Vol. 52 ›› Issue (5): 61-72.DOI: 10.3969/j.issn.1009-086x.2024.05.008

• COMMAND CONTROL AND COMMUNICATION • Previous Articles     Next Articles

Research on Knowledge Graph Construction Method of Early Warning Equipment Based on Deep Learning

Liping YANG1,2, Qiqing FANG1, Yahui HU1, Chenggang GU1, Huimin WANG3   

  1. 1.Air Force Early Warning Academy, Wuhan 430019, China
    2.Zunyi Branch of Guizhou Provincial Corps of the Chinese People's Armed Police Force, Zunyi 563000, China
    3.Guizhou Engineering Vocational College, Tongren 554300, China
  • Received:2023-06-13 Revised:2023-12-18 Online:2024-10-28 Published:2024-11-01

基于深度学习的预警装备知识图谱构建方法研究

杨丽萍1,2, 方其庆1, 胡亚慧1, 谷成刚1, 汪会敏3   

  1. 1.空军预警学院,湖北 武汉 430019
    2.武警贵州省总队遵义支队,贵州 遵义 563000
    3.贵州工程职业学院,贵州 铜仁 554300
  • 作者简介:杨丽萍(1994-),女,贵州遵义人。硕士生,研究方向为预警装备论证、试验与保障指挥。

Abstract:

To fully explore the intrinsic correlation value of massive data, comprehensively and accurately construct the knowledge graph in the field of early warning equipment, a knowledge graph construction method of early warning equipment based on deep learning is proposed. The proposed method takes typical unstructured text data as the research object, constructs the knowledge graph ontology and professional dictionary in the field of early warning equipment, and obtains the early warning equipment knowledge dataset containing 6 468 entity samples and 11 216 relationship samples by driving word segmentation. The entity recognition and relationship extraction are performed based on the knowledge extraction method of multiple deep learning models. The experimental results show that the proposed method has excellent performance in the field of early warning equipment, with the F1 value of the entity recognition model reaching 91.54% and the F1 value of the relationship extraction model reaching 91.05%. The extracted entity relationship triples are stored in the Neo4j graph database, and the early warning equipment field knowledge graph composed of 14 entities and 22 relationships is further constructed and visualized.

Key words: early warning equipment, knowledge graph, deep learning, entity recognition, relationship extraction

摘要:

为了充分挖掘海量数据的内在关联价值,全面准确地构建预警装备领域知识图谱,提出了一种基于深度学习的预警装备知识图谱构建方法。该方法以典型非结构化文本资料为研究对象,构建预警装备领域知识图谱本体和专业词典,以驱动分词得到包含6 468个实体样本和11 216条关系样本的预警装备知识数据集。基于融合多种深度学习模型的知识抽取方法进行实体识别和关系抽取,实验结果表明:所提模型在预警装备领域表现出优异的性能,实体识别模型F1值达到91.54%,关系抽取模型F1值达到91.05%。将提取的实体关系三元组存储在Neo4j图数据库中,进一步构建了由14种实体和22种关系组成的预警装备领域知识图谱并实现可视化。

关键词: 预警装备, 知识图谱, 深度学习, 实体识别, 关系抽取

CLC Number: