知识图谱构建管理系统比较研究与优化构想  

Comparative Study and Optimization Strategies of Knowledge Graph Construction Management Systems

作　　者：马玮璐 鲜国建[1,2] 赵瑞雪 李娇[1,3] 黄永文 孙坦 MA Weilu;XIAN Guojian;ZHAO Ruixue;LI Jiao;HUANG Yongwen;SUN Tan(Agricultural Information Institute,Chinese Academy of Agricultural Sciences,Beijing 100081;Key Laboratory of Agricultural Big Data,Ministry of Agriculture and Rural Affairs,Beijing 100081;Key Laboratory of Knowledge Mining and Knowledge Services in Agricultural Converging Publishing,National Press and Publication Administration,Beijing 100081;Chinese Academy of Agricultural Sciences,Beijing 100081)

机构地区：[1]中国农业科学院农业信息研究所,北京100081 [2]农业农村部农业大数据重点实验室,北京100081 [3]国家新闻出版署农业融合出版知识挖掘与知识服务重点实验室,北京100081 [4]中国农业科学院,北京100081

出　　处：《农业图书情报学报》2023年第4期19-31,共13页Journal of Library and Information Science in Agriculture

基　　金：国家社会科学基金项目“融合多种知识组织体系的认知搜索模式研究”(20BTQ014);国家科技图书文献中心(NSTL)项目“下一代开放知识服务平台关键技术优化集成与系统研发”(2023XM42)。

摘　　要：[目的/意义]知识图谱作为人工智能时代的重要基石,为知识提供了一种新型组织与表示形式,而如何高效构建并合理地管理知识图谱成为当前图谱研究人员的迫切需求。研究聚焦于已有的知识图谱构建管理系统,以期对多款已有系统做全面深入的比较后,总结出当前知识图谱构建管理系统的建设新思路,并为更加通用、实用、好用的知识图谱构建管理系统研发提供参考。[方法/过程]目前大量学者针对知识图谱核心构建流程提出了先进的算法与技术,众多知识图谱相关机构也研发了多种类型的知识图谱构建管理系统,文中选择具有代表性的6款国内外主流知识图谱构建管理系统进行调研,分析各系统在业务流程中的系统特色,在系统的构建流程支持、技术选型及可用性等方面进行总结对比,并围绕当前用户对于知识图谱构建管理系统的最新需求总结已有系统存在的局限。[结果/结论]在深入对比分析的基础上,文中研究了一体化知识图谱协同构建管理系统的建设模式,总结并提出分布式协同构建、多图谱并行管理、多路径知识抽取、多类图存储引擎以及跨媒体与多模态知识图谱等知识图谱构建管理系统建设的优化构想。[Purpose/Significance]Knowledge Graph has become a major research hotspot in the era of artificial intelligence due to its ability to provide a new means of organization and representation of knowledge.As the field continues to evolve,numerous scholars have proposed advanced algorithms and technologies for each core stage of constructing a knowledge graph,and many large domestic and foreign enterprises have also developed their independent knowledge graph management systems.However,the majority of these graph tools developed are designed for commercial use and are often too expensive and difficult to deploy locally for small and medium-sized research teams.This presents a challenge for information organizations such as research libraries with massive resources,which require a more adaptable,universal,and efficient tool to build and manage knowledge graphs.To meet this need,it is important to develop an open-source,user-friendly,and customizable knowledge graph management system that can be easily deployed by small and medium-sized research teams.[Method/Process]In summary,this article offers a thorough and informative analysis of six mainstream knowledge graph management systems,both domestically and internationally.It delves into the unique characteristics of each system within the business process and provides an in-depth comparative analysis based on several important factors,including system functionality,technology selection,open-source availability,and application domains.The article refers to the standard construction process of knowledge graphs and highlights the platform characteristics of each system during the construction process while also examining their limitations based on current data characteristics.In response to practical needs,the article focuses on multi-path,multi-engine,distributed,and collaborative construction,integrating advanced graph algorithms and considering a well-developed underlying graph storage strategy.[Results/Conclusions]As a result,the article presents an in-depth analysis

关 键 词：知识图谱 协同并行 多路径抽取 多图引擎 管理系统 

分 类 号：G251[文化科学—图书馆学]