SIE-GNN: A Link Prediction Algorithm Based on Semantic-Aware Graph Neural Networks

In the field of knowledge graph link prediction, models based on Graph Neural Networks (GNNs) handle multi-faceted and diverse relationships, effectively capturing complex relationships and patterns between entities. However, these models face limitations in addressing complex interactions between entities. A novel link prediction model, SIE-GNN, is proposed to address this issue. This model combines Semantic Evidence-Aware Graph Neural Networks (SE-GNN) with the InteractE model that integrates a triple-attention mechanism. Evaluation results on public datasets indicate that the SIE-GNN model shows improvements in performance metrics compared to the SE-GNN model.