Learning graph attention-aware knowledge graph embedding

doi:10.1016/j.neucom.2021.01.139

CORC > 数学与系统科学研究院 > 中国科学院数学与系统科学研究院 > 应用数学研究所

	Learning graph attention-aware knowledge graph embedding
	Li, Chen 4; Peng, Xutan 3; Niu, Yuhang 1,4; Zhang, Shanghang 6; Peng, Hao 2,4; Zhou, Chuan 5; Li, Jianxin 4
刊名	NEUROCOMPUTING
	2021-10-21
卷号	461 页码:516-529
关键词	Knowledge graph embedding Graph attention mechanism Entity typing Link prediction
ISSN号	0925-2312
DOI	10.1016/j.neucom.2021.01.139
英文摘要	The knowledge graph, which utilizes graph structure to represent multi-relational data, has been widely used in the reasoning and prediction tasks, attracting considerable research efforts recently. However, most existing works still concentrate on learning knowledge graph embeddings straightforwardly and intuitively without subtly considering the context of knowledge. Specifically, recent models deal with each single triple independently or consider contexts indiscriminately, which is one-sided as each knowledge unit (i.e., triple) can be derived from its partial surrounding triples. In this paper, we propose a graph-attention-based model to encode entities, which formulates a knowledge graph as an irregular graph and explores a number of concrete and interpretable knowledge compositions by integrating the graph-structured information via multiple independent channels. To measure the correlation between entities from different angles (i.e., entity pair, relation, and structure), we respectively develop three attention metrics. By making use of our enhanced entity embeddings, we further introduce several improved factorization functions for updating relation embeddings and evaluating candidate triples. We conduct extensive experiments on downstream tasks including entity classification, entity typing, and link prediction to validate our methods. Empirical results validate the importance of our introduced attention metrics and demonstrate that our proposed method can improve the performance of factorization models on large-scale knowledge graphs. (c) 2021 Elsevier B.V. All rights reserved.
资助项目	Natural Science Foundation of China program[U20B2053] ; Natural Science Foundation of China program[62002007] ; Natural Science Foundation of China program[61772151] ; S&T Program of Hebei[20310101D] ; S&T Program of Hebei[SKLSDE-2020ZX-12]
WOS研究方向	Computer Science
语种	英语
出版者	ELSEVIER
WOS记录号	WOS:000704086300003
内容类型	期刊论文
源URL	[http://ir.amss.ac.cn/handle/2S8OKBNM/59347]
专题	应用数学研究所
通讯作者	Li, Chen
作者单位	1.Beihang Univ, State Key Lab Software Dev Environm Lab, Beijing, Peoples R China 2.Beihang Univ, Sch Cyber Sci & Technol, Beijing, Peoples R China 3.Univ Sheffield, Dept Comp Sci, Sheffield, S Yorkshire, England 4.Beihang Univ, Beijing Adv Innovat Ctr Big Data & Brain Comp, Beijing, Peoples R China 5.Chinese Acad Sci, Acad Math & Syst Sci, Beijing, Peoples R China 6.Carnegie Mellon Univ, Dept Elect & Comp Engn, Pittsburgh, PA USA
推荐引用方式 GB/T 7714	Li, Chen,Peng, Xutan,Niu, Yuhang,et al. Learning graph attention-aware knowledge graph embedding[J]. NEUROCOMPUTING,2021,461:516-529.
APA	Li, Chen.,Peng, Xutan.,Niu, Yuhang.,Zhang, Shanghang.,Peng, Hao.,...&Li, Jianxin.(2021).Learning graph attention-aware knowledge graph embedding.NEUROCOMPUTING,461,516-529.
MLA	Li, Chen,et al."Learning graph attention-aware knowledge graph embedding".NEUROCOMPUTING 461(2021):516-529.