Abstract:Knowledge graphs display various entities and their relationships in the real world based on knowledge representation and can analyze and predict the intrinsic relationship between knowledge embeddings through data mining and information processing, which are widely used in search engines, web analytics and smart recommendation areas. As existing knowledge graph information is continuously created and grown, it is a curial task to determine whether the information in knowledge graph is correct and to complete the missing information. In response to this challenge, the researchers proposed a number of graph convolutional network (GCN)-based models to characterize knowledge graphs. The state-of-the-art model is R-GCN, which can effectively extract features. This paper deeply studies the algorithm ideas and results of the R-GCN model, explores whether the model can be further optimized in entity classification and link prediction, and finally, improves the original model and proposes a relational aggregation graph convolutional network. Specifically, this paper finds that a subset of the set of entities may be directly connected to a central entity. All the entities in this subset possess partially identical attributes. At the same time, the relationships between these entities and the central entity may be similar. These similar attributes and relationships can be abstractly aggregated into virtual entities and virtual relationships, respectively, to better extract the topological relationship features. This paper uses the FB15k dataset to evaluate the performance of the proposed model on knowledge graph completion tasks. The experimental results show that the proposed RA-GCN model achieves a certain level improvement compared with the original model and that it can extract knowledge graph topological relationship characteristics more effectively.

RA-GCN

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

Entity Relation Interactive Graph Convolutional Network for Knowledge Embedding

Semantic- and relation-based graph neural network for knowledge graph completion

Rethinking Graph Convolutional Networks in Knowledge Graph Completion

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

Gated Relational Graph Neural Network For Semi-Supervised Learning On Knowledge Graphs

Learnable convolutional attention network for knowledge graph completion

Knowledge Graph Reasoning Based on Attention GCN

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

Graph Neural Network-Based Entity Extraction and Relationship Reasoning in Complex Knowledge Graphs

An efficiency relation-specific graph transformation network for knowledge graph representation learning

A knowledge graph completion model based on contrastive learning and relation enhancement method

Knowledge Graph Representation Learning with Relation-Guided Aggregation and Interaction

Hierarchical Perceptual Graph Attention Network for Knowledge Graph Completion

Text-Enhanced and Relational Context Based Hyperbolic Knowledge Graph Embedding

Relational Graph Neural Network with Hierarchical Attention for Knowledge Graph Completion

Double-Branch Multi-Attention based Graph Neural Network for Knowledge Graph Completion

Cultural Resource Recommender System with Knowledge Graph Convolutional Network.

An attention‐based representation learning model for multiple relational knowledge graph

Multi-Scale Dynamic Convolutional Network for Knowledge Graph Embedding