Enhance prototypical networks with hybrid attention and confusing loss function for few-shot relation classification

doi:10.1016/j.neucom.2022.04.067

	Enhance prototypical networks with hybrid attention and confusing loss function for few-shot relation classification
	Li, Yibing 2,3,4; Ma, Zuchang 2; Gao, Lisheng 2; Wu, Yichen 1,2,3; Xie, Fei 4; Ren, Xiaoye 4
刊名	NEUROCOMPUTING
	2022-07-07
卷号	493
关键词	Relation classification Few-shot learning Hybrid attention Loss BERT
ISSN号	0925-2312
DOI	10.1016/j.neucom.2022.04.067
通讯作者	Ma, Zuchang(zcma@iim.ac.cn)
英文摘要	Relation classification (RC) is a fundamental task to building knowledge graphs and describing semantic formalization. It aims to classify a relation between the head and the tail entities in a sentence. The existing RC method mainly adopts the distant supervision (DS) scheme. However, DS still has the problem of long-tail and suffers from data sparsity. Recently, few-shot learning (FSL) has attracted people's attention. It solves the long-tail problem by learning from few-shot samples. The prototypical networks have a better effect on FSL, which classifies a relation by distance. However, the prototypical networks and their related variants did not consider the critical role of entity words. In addition, not all sentences in support set equally contributed to classifying relations. Furthermore, an entity pair in a sentence may have true and confusing relations, which is difficult for the RC model to distinguish them. A new context encoder BERT_FE is proposed to address those problems, which uses the BERT model as pre-training and fuses the information of head and tail entities by entity word-level attention (WLA). At the same time, the sentence-level attention (SLA) is proposed to give more weight to sentences of the support set similar to the query instance and improve the classification accuracy. A confusing loss function (CLF) is designed to enhance the model's ability to distinguish between true and confusing relations. The experiment results demonstrate that our proposed model (HACLF) is better than several baseline models. (c) 2022 Elsevier B.V. All rights reserved.
资助项目	National Key Research and Development Program of China[2017YFB1002200] ; National Key Research and Development Program of China[2020YFC2005603] ; National Natural Science Foundation of China (NSFC)[61701482] ; Key Projects of the National Natural Science Foundation of Universities in Anhui Province[KJ2020A0112] ; High-level Talents Research Start-up Fund of Hefei Normal University[2020rcjj45] ; Natural Science Foundation of Anhui Province, China[1808085MF191]
WOS研究方向	Computer Science
语种	英语
出版者	ELSEVIER
WOS记录号	WOS:000796189800004
资助机构	National Key Research and Development Program of China ; National Natural Science Foundation of China (NSFC) ; Key Projects of the National Natural Science Foundation of Universities in Anhui Province ; High-level Talents Research Start-up Fund of Hefei Normal University ; Natural Science Foundation of Anhui Province, China
内容类型	期刊论文
源URL	[http://ir.hfcas.ac.cn:8080/handle/334002/130973]
专题	中国科学院合肥物质科学研究院
通讯作者	Ma, Zuchang
作者单位	1.Anhui Jianzhu Univ, Sch Elect & Informat Engn, Hefei 230601, Peoples R China 2.Chinese Acad Sci, Hefei Inst Phys Sci, Inst Intelligent Machines, Anhui Prov Key Lab Med Phys & Technol, Hefei 230031, Peoples R China 3.Univ Sci & Technol China, Sci Isl Branch Grad Sch, Hefei 230026, Peoples R China 4.Hefei Normal Univ, Sch Comp Sci & Technol, Hefei 230601, Peoples R China
推荐引用方式 GB/T 7714	Li, Yibing,Ma, Zuchang,Gao, Lisheng,et al. Enhance prototypical networks with hybrid attention and confusing loss function for few-shot relation classification[J]. NEUROCOMPUTING,2022,493.
APA	Li, Yibing,Ma, Zuchang,Gao, Lisheng,Wu, Yichen,Xie, Fei,&Ren, Xiaoye.(2022).Enhance prototypical networks with hybrid attention and confusing loss function for few-shot relation classification.NEUROCOMPUTING,493.
MLA	Li, Yibing,et al."Enhance prototypical networks with hybrid attention and confusing loss function for few-shot relation classification".NEUROCOMPUTING 493(2022).