Abstract:Graph convolutional networks (GCNs) have received widespread attention in the field of knowledge graph embedding (KGE) due to their powerful graph modeling and neighbor information aggregation capabilities. The GCNs-based embedding methods typically use GCNs to aggregate neighbor entities or relationship information in Euclidean space to enrich the representation of entities and provide initial embeddings for representation learning in KGE methods. However, with the increase of the GCNs convolution layer, the iterative information propagation and feature aggregation may make the entity embeddings in KG become similar, resulting in the inability of the KGE method to distinguish different entities effectively. In addition, embedding hierarchical graph data in Euclidean space will also bring significant distortion. These problems seriously affect the embedded performance of GCNs models. To address these challenges, we propose a hyperbolic graph convolutional network-based knowledge graph embedding (HGCGE) method. Specifically, GCNs convolution operation in hyperbolic space is used to aggregate the adjacent information of entities and capture the deeper relationship between entities, Möbius multiplication and addition are used to embed the entities and relationships to Poincare sphere model to solve the problem of high distortion of embedded representation, and the scoring function based on hyperbolic geometry is designed to increase the attention to entity relationships so as to efficiently identify entities under different relationships. Extensive experiments on four real KG datasets demonstrate that our method outperforms the current strong baseline models on multiple metrics, effectively enhancing the performance of KG embeddings, and it still achieves significant improvements in performance even at low dimensions and low training rounds.

HGCGE: Hyperbolic Graph Convolutional Networks-Based Knowledge Graph Embedding for Link Prediction

Mixed Geometry Message and Trainable Convolutional Attention Network for Knowledge Graph Completion

Hyperbolic Hierarchical Knowledge Graph Embeddings for Link Prediction in Low Dimensions

A Knowledge Graph Embedding Method Based on Neural Network

HyGGE: Hyperbolic Graph Attention Network for Reasoning over Knowledge Graphs

Text-Enhanced and Relational Context Based Hyperbolic Knowledge Graph Embedding

Hierarchical Embedding Methods for Large-scale Complex Knowledge Graphs

Knowledge Association with Hyperbolic Knowledge Graph Embeddings

A data-centric framework of improving graph neural networks for knowledge graph embedding

Entity Relation Interactive Graph Convolutional Network for Knowledge Embedding

Convolutional Models with Multi-Feature Fusion for Effective Link Prediction in Knowledge Graph Embedding

Geometry Interaction Knowledge Graph Embeddings

Meta-Knowledge Transfer for Inductive Knowledge Graph Embedding

Knowledge graph embedding based on dynamic adaptive atrous convolution and attention mechanism for link prediction

CoPE: Composition-based Poincaré embeddings for link prediction in knowledge graphs

Multi-Scale Dynamic Convolutional Network for Knowledge Graph Embedding

Rethinking Graph Convolutional Networks in Knowledge Graph Completion

Efficiently Embedding Dynamic Knowledge Graphs

RIECN: Learning Relation-Based Interactive Embedding Convolutional Network for Knowledge Graph.

Link Prediction in Knowledge Graphs Based on Hyperbolic Graph Attention Networks

Learning higher-order features for relation prediction in knowledge hypergraph