面向知识图谱补全的归纳学习研究综述  

Survey on Inductive Learning for Knowledge Graph Completion

作　　者：梁新雨 司冠南[1] 李建辛 田鹏新 安兆亮 周风余[2] LIANG Xinyu;SI Guannan;LI Jianxin;TIAN Pengxin;AN Zhaoliang;ZHOU Fengyu(School of Information Science&Electrical Engineering,Shandong Jiaotong University,Jinan 250357,China;School of Control Science&Engineering,Shandong University,Jinan 250000,China)

机构地区：[1]山东交通学院信息科学与电气工程学院,济南250357 [2]山东大学控制科学与工程学院,济南250000

出　　处：《计算机科学与探索》2023年第11期2580-2604,共25页Journal of Frontiers of Computer Science and Technology

基　　金：国家自然科学基金(61375084);山东省自然科学基金(ZR2019MF064)。

摘　　要：知识图谱补全能够使知识图谱更加完整。然而,传统的知识图谱补全方法假定在测试时所有实体和关系都出现在训练过程,由于现实世界知识图谱的演变性质,一旦出现不可见实体或不可见关系,就需要重新训练知识图谱。面向知识图谱补全的归纳学习旨在补全包含不可见实体或不可见关系的三元组,而无需从头开始训练知识图谱,因此近年来受到广泛关注。首先从知识图谱的基本概念出发,将知识图谱补全分为两大类,直推式和归纳式;其次从归纳式的知识图谱补全的理论角度出发,分为半归纳和全归纳这两类,并从该角度对模型进行总结归纳;然后从归纳式的知识图谱补全的技术角度出发,分为基于结构信息和基于额外信息这两大类,将基于结构信息的方法细分为基于归纳嵌入、基于逻辑规则和基于元学习这三类,将基于额外信息的方法细分为基于文本信息和其他信息这两类,并对当下方法进一步深入细分、分析和对比;最后对未来的研究方向进行展望。Knowledge graph completion can make knowledge graph more complete.However,traditional knowledge graph completion methods assume that all test entities and relations appear in the training process.Due to the evolving nature of real world KG,once unseen entities or relations appear,the knowledge graph needs to be retrained.Inductive learning for knowledge graph completion aims to complete triples containing unseen entities or unseen relations without training the knowledge graph from scratch,so it has received much attention in recent years.Firstly,starting from the basic concept of knowledge graph,this paper divides knowledge graph completion into two categories:transductive and inductive.Secondly,from the theoretical perspective of inductive knowledge graph completion,it is divided into two categories:semi-inductive and fully-inductive,and the models are summarized from this perspective.Then,from the technical perspective of inductive knowledge graph completion,it is divided into two categories:based on structural information and based on additional information.The methods based on structural information are subdivided into three categories:based on inductive embedding,based on logical rules and based on meta learning,and the methods based on additional information are subdivided into two categories:based on text information and other information.The current methods are further subdivided,analyzed and compared.Finally,it forecasts the main research directions in the future.

关 键 词：知识图谱 知识图谱补全 归纳学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]