Supervised Entity Alignment

Supervised entity alignment (EA) tackles the problem of identifying corresponding entities across different knowledge graphs (KGs), aiming to link entries representing the same real-world object. Recent research heavily focuses on improving EA's performance with limited labeled data, exploring techniques like pseudo-labeling, mixture teaching, and self-supervised learning within frameworks employing graph neural networks, optimal transport, and Gromov-Wasserstein distances. These advancements are crucial for integrating and harmonizing diverse KGs, enabling more comprehensive knowledge discovery and facilitating applications in areas like data integration, cross-lingual information retrieval, and question answering.