知识图谱表示学习技术综述  

Survey of Knowledge Graph Representation Learning Technology

作　　者：石昌友 夏榕泽 黄蔚 韩欢 周静 SHI Changyou;XIA Rongze;HUANG Wei;HAN Huan;ZHOU Jing(Communication Sergeant School,Army Engineering University of PLA,Chongqing 400035,China)

机构地区：[1]陆军工程大学通信士官学校,重庆400035

出　　处：《计算机测量与控制》2025年第3期1-11,29,共12页Computer Measurement &Control

摘　　要：知识图谱表示学习,是将知识图谱中实体与关系以低维稠密向量表示的技术,在知识图谱驱动的人工智能研究中发挥着基础性支撑作用,已是当下研究热点,引起学者广泛关注,并取得很多研究成果;从表示学习的基本概念出发,系统性地阐述知识图谱表示学习方法最新研究进展,具体从算法模型的问题背景、算法模型原理、算法模型特点等方面进行详细论述;聚焦平移模型类算法,将模型算法细分成:单数据空间、多数据空间、概率空间、外部信息融合等类型,详细分析代表性模型,并梳理各算法间演化关系,从定量和定性两个维度归纳总结平移类算法模型;从表示空间类型、编码模型、外部信息融合、实时知识表示学习等方面展望未来发展趋势。Knowledge graph learning is a technique in which entity and relation in knowledge graph are represented by a low dimensional dense vector,which plays a significant role in the research of knowledge graph driven artificial intelligence.It is a current research hotspot and attracts widespread attention from scholar,achieving many research results.From basic concept in knowledge representation learning,this paper systematically elaborates on the latest research progress in knowledge representation learning methods,specifically discussing the background,principle,and characteristics of algorithm models in detail.By focusing on translation-based models,the algorithms are subdivided into the types of single data space,multiple data space,probability space,and external information fusion,implementing the detailed analysis of representative models and sorting out the evolutionary relationships between various algorithms,and the translation algorithms are summarized from quantitative and qualitative dimensions.Finally,future development trends for knowledge representation learning are expected from aspects such as representation space types,encoding models,external information fusion,real-time knowledge representation learning.

关 键 词：知识图谱 知识表示学习 平移模型 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]