Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (18): 59-73.DOI: 10.3778/j.issn.1002-8331.2212-0119

• Research Hotspots and Reviews • Previous Articles     Next Articles

Research Progress of Knowledge Graph Completion Based on Knowledge Representation Learning

YU Mengbo, DU Jianqiang, LUO Jigen, NIE Bin, LIU Yong, QIU Junyang   

  1. College of Computer Science, Jiangxi University of Traditional Chinese Medicine, Nanchang 330004, China
  • Online:2023-09-15 Published:2023-09-15



  1. 江西中医药大学 计算机学院,南昌 330004

Abstract: A knowledge graph(KG) is a graph-based data structure in which knowledge is presented in the form of triples(head entities, relationships, tail entities). With the development of artificial intelligence, knowledge graph has played an important role in intelligent system recommendation, question answering, knowledge search and so on. However, the constructed knowledge graph is incomplete, which affects the downstream task application of knowledge graph. Knowledge graph completion can solve this problem well. In recent years, knowledge graph completion method based on knowledge representation learning has become a research hotspot. It learns the embedded features of entities and relations in low-dimensional continuous vector space in the form of representation vector, aiming at predicting unknown factual information for knowledge graph completion. According to the different types of KG, firstly, it is divided into static knowledge graph completion, temporal knowledge graph completion and multimodal knowledge graph completion. Secondly, the key problems, design ideas, model evaluation and other aspects intended to be solved by these three knowledge graph completion methods are compared and summarized. Finally, the future development direction of knowledge graph completion is prospected to provide references for researchers in related fields.

Key words: knowledge graph, knowledge graph completion, knowledge representation learning

摘要: 知识图谱(KG)是一种基于图的数据结构,其知识是以三元组的形式呈现,即(头实体,关系,尾实体)。随着人工智能的发展,知识图谱已在系统推荐、智能问答、知识搜索等领域发挥了重要作用。然而构建的知识图谱具有不完整性,影响了知识图谱的下游任务应用,知识图谱补全能够很好地解决这一问题。近年来,基于知识表示学习的知识图谱补全方法成为研究的热点,其以表示向量的形式在低维连续向量空间中学习实体和关系的嵌入特征,旨在预测未知的事实信息进行知识图谱补全。根据KG类型的不同,将其分为静态知识图谱补全、时序知识图谱补全以及多模态知识图谱补全,对这三类知识图谱补全方法拟解决的关键问题、设计思路、模型评价等方面进行对比总结,展望知识图谱补全未来的发展方向,为相关领域的研究人员提供参考。

关键词: 知识图谱, 知识图谱补全, 知识表示学习