面向试验数据的标签本体及标签实例推荐方法  

A Recommendation Method Based on Tags and Tag Instances for Test Data

作　　者：张骁雄 张明星[1,2,3] 谢志豪 任智颖 ZHANG Xiaoxiong;ZHANG Mingxing;XIE Zhihao;REN Zhiying(The Sixty-Third Research Institute,National University of Defense Technology,Nanjing 210007,China;Laboratory for Big Data and Decision,National University of Defense Technology,Changsha 410073,China;School of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China;China Ordnance Industry Computer Application Technology Institute,Beijing 100089,China)

机构地区：[1]国防科技大学第六十三研究所,南京210007 [2]国防科技大学大数据与决策实验室,长沙410073 [3]南京信息工程大学电子与信息工程学院,南京210044 [4]中国兵器工业计算机应用技术研究所,北京100089

出　　处：《火力与指挥控制》2023年第10期41-48,共8页Fire Control & Command Control

基　　金：特殊领域青年人才托举工程项目(2021-JCJQ-QT-050);国防科技大学校科研计划项目(ZK20-46)。

摘　　要：针对海量试验数据标签管理及标签管理智能化问题,提出一种标签本体及标签实例推荐方法。采集结构化、非结构化数据以及半结构化数据,根据试验数据类型,构建本体概念,形成标签库;对图像、文本、音频、视频、纸质等多模态数据采用不同方法进行装备实体以及实体关系抽取,构建标签实例库;利用规则映射和自然语言处理方法对标签本体与标签实例进行关系映射;最后,挖掘用户个人信息以及使用标签信息,结合个人信息以及标签信息,形成基于标签及标签实例的智能推荐。对比现有模型,该推荐模型在MAE和MSE指标上分别降低了8.82%和5.56%,AUC指标提高了13.33%,对试验数据智能化管理具有重要意义。Aiming at the problems of tag management of massive test data and intelligence of tag management,this paper proposes a method of tags and tag instances recommendation.The method includes collecting structured data,unstructured data and semi-structured data,building ontology concepts based on the type of experimental data,and forming a tag library.Then,the multi-modal data of image,text,audio,video,and paper are extracted by different methods,and the tag instance library is constructed.In addition,the relationships between tags and tag instances are mapped by using rule mapping and natural language processing.At last,the personal information and the tag information are combined by mining the user's personal information and using the tag information and forming intelligent recommendations based on tags and tag instances.Compared with the existing model,the MAE and MSE indicators of the model are reduced by 8.82%and 5.56%respectively,and the AUC index is increased by 13.33%,which is of great significance for the intelligent management of test data.

关 键 词：试验数据 标签本体 多模态数据 自然语言处理 智能推荐 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]