EntityResolutionWithPyG

EmbDI is a framework that uses a graph as an intermediate representation for relational tables to generate embeddings of schema elements such as tuples, tokens, and columns. It was proved that this approach causes a lower loss of information if compared to frameworks that consider tuples as sentences. In this project, I re-implemented from scratch EmbDI using PyTorch geometric to generate the embeddings. My objective was to test the performances of this deep learning library based on Graph Neural Networks on the entity resolution task.