A Multi-Modal Entity Alignment Method with Inter-Modal Enhancement

Due to inter-modal effects hidden in multi-modalities and the impact of weak modalities on multi-modal entity alignment, a Multi-modal Entity Alignment Method with Inter-modal Enhancement (MEAIE) is proposed. This method introduces unique modality called numerical modal aspect applies feature encoder encode it. In embedding stage, this paper utilizes visual features enhance relation representation influence attribute weight distribution. Then, attention layers contrastive learning strengthen mitigate modalities. order evaluate performance proposed method, experiments are conducted three public datasets: FB15K, DB15K, YG15K. By combining datasets pairs, compared current state-of-the-art alignment models, model achieves 2% 3% improvement Top-1 Hit Rate(Hit@1) Mean Reciprocal Rank (MRR), demonstrating its feasibility effectiveness.