应用知识图谱的推荐方法与系统  

Recommendation Methods and Systems Using Knowledge Graph

作　　者：饶子昀 张毅 刘俊涛[1] 曹万华[1] RAO Zi-Yun;ZHANG Yi;LIU Jun-Tao;CAO Wan-Hua(Wuhan Digital Engineering Institute,Wuhan 430205)

机构地区：[1]武汉数字工程研究所,武汉430205

出　　处：《自动化学报》2021年第9期2061-2077,共17页Acta Automatica Sinica

基　　金：国家自然科学基金(61403350)资助。

摘　　要：数据稀疏和冷启动是当前推荐系统面临的两大挑战.以知识图谱为表现形式的附加信息能够在某种程度上缓解数据稀疏和冷启动带来的负面影响,进而提高推荐的准确度.本文综述了最近提出的应用知识图谱的推荐方法和系统,并依据知识图谱来源与构建方法、推荐系统利用知识图谱的方式,提出了应用知识图谱的推荐方法和系统的分类框架,进一步分析了本领域的研究难点.本文还给出了文献中常用的数据集.最后讨论了未来有价值的研究方向.Data sparsity and cold-start problems are two major challenges for current recommendation systems.The additional information in the form of a knowledge graph can alleviate these challenges to a certain extent,and integrating this information into recommender systems can improve the accuracy of the recommendation.This paper reviews recommendation methods and systems using knowledge graphs proposed recently,and proposes a classification framework for this kind of recommendation methods according to the source and construction methods of knowledge graphs and the way the recommendation systems use knowledge graph.We further analyze the research difficulties in this field.We also present commonly used datasets in the literature.Finally,future valuable research directions are discussed.

关 键 词：推荐系统 协同过滤 知识图谱 表示学习 知识推理 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]