结合GAT与卷积神经网络的知识超图链接预测

doi:10.3778/j.issn.1002-8331.2401-0443

摘要/Abstract

摘要： 知识超图（knowledge hypergraph，KHG）是一种超图结构的知识图谱。知识超图链接预测是基于已知的实体和关系来预测缺失的实体或关系，具有重要的意义和价值。然而，现有基于神经网络的知识超图链接预测方法，只关注关系事实局部的语义特征，缺乏对关系事实之间关联特征的表示学习。针对以上问题，提出了一种基于图注意力网络与卷积神经网络的链接预测方法（knowledge prediction based on GAT and convolutional neural network，HPGC）。一方面，采用改进的卷积网络（convolutional neural network，CNN）提取知识超图中节点实体表示的局部特征；另一方面，使用改进的GAT对节点和关系进行注意力建模，捕获节点之间的全局特征关系，并将两者进行融合，从而获取关系事实更全面的邻域结构，丰富超图关系事实的语义表示。此外，针对HPGC的GAT层输出矢量问题，引入多层感知机（multilayer perceptron，MLP）和正则化技术，提高模型训练的泛化能力。真实数据集上的大量实验结果验证了所提出方法的预测性能均优于基线方法。

关键词: 知识超图, 链接预测, 卷积神经网络, 注意力机制

Abstract: Knowledge hypergraph (KHG) is a hypergraph structured knowledge graph. Knowledge hypergraph link prediction is based on known entities and relationships to predict missing entities or relationships, which is of great significance and value. However, the existing neural network-based methods of knowledge hypergraph link prediction only focus on the semantic features localized to the relationship facts and lack the representation learning of the association features between the relationship facts. To address the above problems, a knowledge hypergraph link prediction method based on graph attention network and convolutional neural network (HPGC) is proposed. On the one hand, an improved convolutional network is used to extract local features of node entity representations in the knowledge hypergraph; on the other hand, an improved GAT is used for attention modeling of nodes and relations, capturing global feature relations between nodes and merging them, so as to obtain a more comprehensive neighborhood structure of the relation facts, and to enrich the semantic representations of the relation facts in the hypergraph. In addition, to address the GAT layer output vector problem of HPGC, which introduces a multilayer perceptron (MLP) and regularization techniques to enhance the generalization ability of model training. Extensive experimental results on real datasets validate that the prediction performances of the proposed methods outperform the baseline methods.

Key words: knowledge hypergraph, link prediction, convolutional neural network, attention mechanism

庞俊, 马志芬, 林晓丽, 王蒙湘. 结合GAT与卷积神经网络的知识超图链接预测[J]. 计算机工程与应用, 2025, 61(9): 194-201.

PANG Jun, MA Zhifen, LIN Xiaoli, WANG Mengxiang. Knowledge Hypergraph Link Prediction Based on GAT and Convolutional Neural Network[J]. Computer Engineering and Applications, 2025, 61(9): 194-201.

参考文献

[1] 张天成, 田雪, 孙相会, 等. 知识图谱嵌入技术研究综述[J]. 软件学报, 2023, 34(1): 277-311.
ZHANG T C, TIAN X, SUN X H, et al. Overview on knowledge graph embedding technology research[J]. Journal of Software, 2023, 34(1): 277-311.
[2] WEN J, LI J, MAO Y, et al. On the representation and embedding of knowledge bases beyond binary relations[J]. arXiv:1604.08642, 2016.
[3] FATEMI B, TASLAKIAN P, VáZQUEZ D, et al. Knowledge hypergraphs: prediction beyond binary relations[C]//Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence, 2020: 2191-2197.
[4] 王永贵, 陈书铭, 刘义海, 等. 结合超图对比学习和关系聚类的知识感知推荐算法[J]. 计算机科学与探索, 2024, 18(8): 2140-2155.
WANG Y G, CHEN S M, LIU Y H, et al.?Knowledge-aware recommendation algorithm combining hypergraph contrast learning and relational clustering[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(8): 2140-2155.
[5] HAN J, CHENG B, WANG X. Two-phase hypergraph based reasoning with dynamic relations for multi-hop KBQA[C]//Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, 2020: 3615-3621.
[6] BURKE R. Knowledge-based recommender systems[J]. Encyclopedia of Library and Information Systems, 2000, 69(32): 175-186.
[7] WANG Z, ZHANG J, FENG J, et al. Knowledge graph embedding by translating on hyperplanes[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2014.
[8] ZHANG R, LI J, MEI J, et al. Scalable instance reconstruction in knowledge bases via relatedness affiliated embedding[C]//Proceedings of the 2018 World Wide Web Conference, 2018: 1185-1194.
[9] LIU Y, YAO Q, LI Y. Generalizing tensor decomposition for N-ary relational knowledge bases[C]//Proceedings of the Web Conference 2020, 2020: 1104-1114.
[10] 王培妍, 段磊, 郭正山, 等. 基于张量分解的知识超图链接预测模型[J]. 计算机研究与发展, 2021, 58(8):1599-1611.
WANG P Y, DUAN L, GUO Z S, et al. Knowledge hypergraph link prediction model based on tensor decomposition[J]. Journal of Computer Research and Development, 2021, 58(8):1599-1611.
[11] PANG J, QIN H C, LIU Y, et al. Two birds with one stone: a link prediction model for knowledge hypergraph based on fully-connected tensor decomposition[C]//Proceedings of the International Conference on Advanced Data Mining and Applications. Cham: Springer Nature Switzerland, 2023: 78-90.
[12] GUAN S, JIN X, WANG Y, et al. Link prediction on N-ary relational data[C]//Proceedings of the World Wide Web Conference, 2019: 583-593.
[13] GUAN S, JIN X, GUO J, et al. Neuinfer: knowledge inference on n-ary facts[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 2020: 6141-6151.
[14] ROSSO P, YANG D, CUDRé-MAUROUX P. Beyond triplets: hyper-relational knowledge graph embedding for link prediction[C]//Proceedings of the Web Conference 2020, 2020: 1885-1896.
[15] GUAN S, JIN X, GUO J, et al. Link prediction on N-ary relational data based on relatedness evaluation[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(1): 672-685.
[16] CHEN Z, WANG X, WANG C, et al. Explainable link prediction in knowledge hypergraphs[C]//Proceedings of the 31st ACM International Conference on Information & Knowledge Management, 2022: 262-271.
[17] HUANG J, LEI F, JIANG J, et al. Multi-order hypergraph convolutional networks integrated with self-supervised learning[J]. Complex & Intelligent Systems, 2023, 9(4): 4389-4401.
[18] VELI?KOVI? P, CUCURULL G, CASANOVA A, et al. Graph attention networks[J]. arXiv:1710.10903, 2017.
[19] 张文豪, 徐贞顺, 刘纳, 等. 知识图谱补全方法研究综述[J]. 计算机工程与应用, 2024, 60(12): 61-73.
ZHANG W H, XU Z S, LIU N, et al. Overview of knowledge graph completion methods[J]. Computer Engineering and Applications, 2024, 60(12): 61-73.
[20] KAZEMI S M, POOLE D. SimplE embedding for link prediction in knowledge graphs[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems (NIPS’18), Montréal, Canada. Cambridge: MIT Press, 2018: 4289-4300.
[21] WANG G, ZENG Y, LI R H, et al. Temporal graph cube[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(12):13015-13030.
[22] GU Y, YU K, SONG Z, et al. Distributed hypergraph processing using intersection graphs[J]. IEEE Transactions on Knowledge and Data Engineering, 2022, 34(7):3182-3195.
[23] YU D, GU Y, XIONG C, et al. CompleQA: benchmarking the impacts of knowledge graph completion methods on question answering[C]//Findings of the Association for Computational Linguistics: EMNLP 2023, 2023: 12748-12755.
[24] KADLEC R, BAJGAR O, KLEINDIENST J. Knowledge base completion: baselines strike back[J]. arXiv:1705.10744, 2017.