Few-Shot Relation Prediction of Knowledge Graph via Convolutional Neural Network with Self-Attention

Data Science and Engineering

Research Paper

Zhengbao Sun¹  Kun Yue²  Jiahui Wang²  Shanna Zhong²  Liang Duan²  Yan Fang³ 

[1] School of Engineering, Yunnan University, 650500, Kunming, China;School of Information Science and Engineering, Yunnan University, 650500, Kunming, China;Yunnan Key Laboratory of Intelligent Systems and Computing, Yunnan University, 650500, Kunming, China;Yunnan Key Laboratory of Intelligent Systems and Computing, Yunnan University, 650500, Kunming, China;

关键词: Knowledge graph;    Few-shot relation prediction;    Convolutional neural network;    Self-attention;   

DOI  :  10.1007/s41019-023-00230-x

 received in 2023-05-18, accepted in 2023-08-28,  发布年份 2023

来源: Springer

【 摘 要 】

Knowledge graph (KG) has become the vital resource for various applications like question answering and recommendation system. However, several relations in KG only have few observed triples, which makes it necessary to develop the method for few-shot relation prediction. In this paper, we propose the Convolutional Neural Network with Self-Attention Relation Prediction (CARP) model to predict new facts with few observed triples. First, to learn the relation property features, we build a feature encoder by using the convolutional neural network with self-attention from the few observed triples rather than background knowledge. Then, by incorporating the learned features, we give an embedding network to learn the representation of incomplete triples. Finally, we give the loss function and training algorithm of our CARP model. Experimental results on three real-world datasets show that our proposed method improves Hits@10 by 48% on average over the state-of-the-art competitors.

【 授权许可】

CC BY   
© The Author(s) 2023

【 图 表 】

Fig. 9

Fig. 6

Fig. 1

Fig. 2

Fig. 1

12951_2016_171_Article_IEq1.gif

12936_2023_4742_Article_IEq36.gif

Fig. 5

Fig. 5

Fig. 3

Fig. 2

Fig. 3

Fig. 1

Fig. 5

12951_2017_252_Article_IEq1.gif

Fig. 4

Fig. 2

Fig. 2

12936_2023_4742_Article_IEq20.gif

Fig. 1

12936_2023_4742_Article_IEq18.gif

Fig. 2

Fig. 5

Fig. 2

Fig. 3

Fig. 4

Fig. 6

Fig. 4

Fig. 3

Fig. 3

Fig. 7

Fig. 1

Fig. 1

12936_2023_4742_Article_IEq9.gif

Fig. 6

Fig. 2

Fig. 4

Fig. 8

Fig. 2

Fig. 5

Fig. 4

12936_2016_1411_Article_IEq110.gif

Fig. 3

Fig. 1

Fig. 2

Fig. 4

Fig. 2

12888_2023_5283_Article_IEq2.gif

12888_2023_5283_Article_IEq1.gif

Fig. 1

12936_2017_1963_Article_IEq54.gif

Fig. 3

Fig. 1

12936_2015_1050_Article_IEq030.gif

Fig. 2

12936_2017_2118_Article_IEq1.gif

12951_2017_255_Article_IEq56.gif

Fig. 2

Fig. 3

Fig. 1

Fig. 5

Fig. 14

Fig. 11

12936_2017_1882_Article_IEq9.gif

Fig. 7

Fig. 3

12951_2017_283_Article_IEq3.gif

Fig. 13

Fig. 4

Fig. 3

Fig. 2

Fig. 7

12951_2017_255_Article_IEq54.gif

Fig. 5

12951_2017_255_Article_IEq53.gif

Fig. 2

12888_2023_5292_Article_IEq1.gif

Fig. 1

Fig. 1

Fig. 2

12951_2016_246_Article_IEq16.gif

12951_2016_246_Article_IEq15.gif

Fig. 6

12951_2016_246_Article_IEq13.gif

12951_2016_246_Article_IEq12.gif

Fig. 1

Fig. 1

Fig. 9

Fig. 4

12951_2017_255_Article_IEq52.gif

12951_2017_255_Article_IEq51.gif

Fig. 4

Fig. 2

Fig. 1

Fig. 1

Fig. 4

Fig. 10

Fig. 1

Fig. 1

Fig. 2

Scheme 1

12951_2016_223_Article_IEq1.gif

12951_2017_255_Article_IEq50.gif

12951_2017_255_Article_IEq49.gif

12936_2017_1904_Article_IEq1.gif

Fig. 1

Fig. 5

Fig. 6

42004_2023_1031_Figa_HTML.png

12951_2016_246_Article_IEq9.gif

42004_2023_1031_Article_IEq16.gif

12951_2016_246_Article_IEq8.gif

Fig. 5

Fig. 6

Fig. 1

Fig. 1

12951_2017_255_Article_IEq48.gif

Fig. 3

Fig. 5

Fig. 1

Fig. 2

Fig. 10

Fig. 1

Fig. 4

Fig. 2

Fig. 8

12951_2016_246_Article_IEq7.gif

Fig. 1

12951_2016_246_Article_IEq6.gif

Fig. 8

40538_2023_473_Article_IEq1.gif

12951_2015_155_Article_IEq78.gif

Fig. 7

Fig. 8

Fig. 4

12951_2015_155_Article_IEq77.gif