基于实体关系抽取的军事装备图谱构建  

Military equipment knowledge graph construction based on entity relationship extraction

作　　者：王依科 吴振乾 WANG Yike;WU Zhenqian(Jiangsu Automation Research Institute,Lianyungang 222000,China)

机构地区：[1]江苏自动化研究所,江苏连云港222000

出　　处：《现代电子技术》2024年第15期163-168,共6页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(U21A20488)。

摘　　要：由于信息技术的不断进步,许多军事装备数据库结构松散,难以有效利用,导致效率低下、管理混乱等问题。针对上述问题,提出一种基于CRF和句法分析树的实体关系提取方法。通过海量数据训练,优化军事知识图谱构建方法,将单算法提取方法改进为三元数据提取方法,完成军事装备图谱构建。实验结果表明,该方法准确率可达72%,且加入置信模型后,准确率提高了12.6%,综合评价准确率可达78.11%。这一结果对军事装备领域知识图谱的构建具有重要的实用价值。Because of the continuous advancement of information technology,it is difficult to utilize many military equipment databases effectively due to their incompact structures,which results in low efficiency and chaotic management.In view of the above,an entity relationship extraction method based on CRF(conditional random field)and syntax analysis tree is proposed.The construction method of military knowledge graph is optimized by the training of massive data,and the single algorithm extraction method is changed into a three element extraction method,so as to complete the construction of military equipment graph.The experimental results show that the accuracy of the method can reach 72%.After adding the confidence model,its accuracy is increased by 12.6%,and its comprehensive evaluation accuracy can reach 78.11%.This result has important practical value for the construction of knowledge graphs in the field of military equipment.

关 键 词：军事装备 关系抽取 知识图谱 数据库结构 置信模型 三元数据提取 

分 类 号：TN919-34[电子电信—通信与信息系统] TP399[电子电信—信息与通信工程]