垂直领域知识图谱构建及应用平台的设计与实现  

Design and Implementation of a Platform for Domain Knowledge Graph Construction and Applications

作　　者：张博尧[1,2] 曹荣强 万萌[1] 孙境棋 王彦棡 王珏 赵永华[1,2] ZHANG Boyao;CAO Rongqiang;WAN Meng;SUN Jingqi;WANG Yangang;WANG Jue;ZHAO Yonghua(Department of Artificial Intelligence Technology and Application Development,Computer Network Information Center,Chinese Academy of Sciences,Beijing 100083,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院计算机网络信息中心,人工智能技术与应用发展部,北京100083 [2]中国科学院大学,北京100049

出　　处：《数据与计算发展前沿》2023年第3期111-122,共12页Frontiers of Data & Computing

基　　金：中国国家电网有限公司总部管理科技项目“自主可控电力人工智能开放平台关键技术研究”(5700-20215-8261A-0-0-00)。

摘　　要：【目的】知识图谱可以将客观世界的实体、概念、关系等抽象成图数据结构,是垂直领域建模和应用的研究热点,本文研发和实现了从资源调度、知识图谱构建到应用场景全过程的自动化和一体化平台。【方法】以微服务架构的方式部署了数据处理、算法仓库及知识图谱公共组件,采用了低耦合、高内聚和多层次的架构以提升平台处理知识抽取、存储与应用等不同任务时多模块间的协作能力,将计算资源虚拟化以提升初始化、生产与日常维护等平台的不同使用阶段的资源调度能力。【结果】建成和部署了垂直领域知识图谱构建及应用平台,提供了数据处理、自动化知识抽取、图计算及扩展、图表示学习及应用等功能。以金融领域产业链知识图谱的构建及表示学习应用的过程为例,验证了平台的可用性。【结论】该平台实现了垂直领域知识图谱从构建、图表示学习到领域应用的全流程覆盖,有效地降低了用户的使用难度,提高了资源的利用效率,也可以扩展到其它垂直领域。ed into a graph data structure by knowledge graph technology.It is a research hotspot of modeling and applications in the vertical domain.In this paper,an automated and integrated platform is developed and implemented,which can cover the whole process from resource scheduling and knowledge graph construction to applications.[Methods]The data processing module,algorithm warehouse,and knowledge graph common components are deployed in the form of microservice architecture.In order to improve the cooperation ability between multiple modules,the platform adopts a low-coupling,high-cohesion,and multi-layered architecture to handle different tasks such as knowledge extraction,storage,and applications.[Results]The knowledge graph construction and application platform in the vertical domain are built and deployed.It provides functions such as data processing,automatic knowledge extraction,graph computation and extension,graph representation learning and application.The usability of the platform is verified,with the construction of a knowledge graph and application of representation learning as a typical case in the financial domain.[Conclusions]The platform covers the whole process from graph construction,graph representation learning to domain applications for knowledge graphs in the vertical domain.It effectively reduces the difficulty of use,improves the efficiency of resource utilization,and can be extended to other vertical domains.

关 键 词：知识图谱 资源虚拟化 图表示学习 微服务技术 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]