Unsupervised Entity Alignment Model via Optimal Transport
Zhao Xin (1811336)
The task of entity alignment aims at finding corresponding entities that have the same real-world semantics across different knowledge graphs. Recently, embedding-based methods have been proposed for this task. Such models assume that a large number of aligned entities are known. However, in real-world dataset, such supervised data are difficult to obtain. To tackle this problem, we explore unsupervised methods by modeling entity alignment as an optimal transport problem and propose
a model using Gromov-Wasserstein distance. In our experiment, we demonstrate that our model can achieve good performance.