《计算机应用研究》|Application Research of Computers

知识表示学习方法研究综述

Survey of knowledge representation learning methods

免费全文下载 (已被下载 次)  
获取PDF全文
作者 张正航,钱育蓉,行艳妮,赵鑫
机构 新疆大学 a.软件学院;b.新疆维吾尔自治区信号检测与处理重点实验室,乌鲁木齐 830046
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)04-001-0961-07
DOI 10.19734/j.issn.1001-3695.2020.04.0094
摘要 近年来,知识表示学习已经成为知识图谱领域研究的热点。为了及时掌握当前知识表示学习方法的研究现状,通过归纳与整理,将具有代表性的知识表示方法进行了介绍和归类,主要分为传统的知识表示模型、改进的知识表示模型、其他的知识表示模型。对每一种方法解决的问题、算法思想、应用场景、评价指标、优缺点进行了详细归纳与分析。通过研究发现,当前知识表示学习主要面临关系路径建模、准确率、复杂关系处理的挑战。针对这些挑战,展望了采用关系的语义组成来表示路径、采用实体对齐评测指标、在实体空间和关系空间建模,以及利用文本上下文信息以扩展KG的语义结构的解决方案。
关键词 知识图谱; 知识表示学习; 实体对齐; 链接预测; 三元组分类
基金项目 国家自然科学基金资助项目(61966035)
新疆维吾尔自治区智能多模态信息处理团队(XJEDU2017T002)
新疆维吾尔自治区研究生创新项目(XJ2019G072)
本文URL http://www.arocmag.com/article/01-2021-04-001.html
英文标题 Survey of knowledge representation learning methods
作者英文名 Zhang Zhenghang, Qian Yurong, Xing Yanni, Zhao Xin
机构英文名 a.College of Software,b.Key Laboratory of Signal Detection & Processing in Xinjiang Uygur Autonomous Region,Xinjiang University,Urumqi 830046,China
英文摘要 In recent years, knowledge representation learning has become a hot topic in the field of knowledge graph. In order to grasp the current research status of knowledge representation learning methods in time, this paper introduced and classified the representative knowledge representation methods through induction and sorting, which were mainly divided into traditional knowledge representation model, improved knowledge representation model and other knowledge representation models. This paper summarized and analyzed the problems, algorithm ideas, application scenarios, evaluation indicators, advantages and disadvantages of each method in detail. Through research, this paper found that the current knowledge representation learning mainly faces the challenges of relationship path modeling, accuracy and complex relationship processing. Aiming at these challenges, this paper looked forward to using semantic composition of relationship to represent path, using entity alignment evaluation index, modeling in entity space and relationship space, and using text context information to expand the solution of KG's semantic structure resolution.
英文关键词 knowledge graph; knowledge representation learning; entity alignment; link prediction; triple classification
参考文献 查看稿件参考文献
 
收稿日期 2020/4/26
修回日期 2020/6/8
页码 961-967
中图分类号 TP182
文献标志码 A