Efficient Table Embeddings via Self-Supervised Structural-Semantic Graph Autoencoder

Jinlong Tian, Shixuan Liu, Ruochun Jin, Mengmeng Li, Yanfang Zhou, Xinhai Xu, Yuhua Tang

Abstract: Highlights•We introduce TEA, a framework for learning embeddings from tabular data.•Our TEA framework models complex structural and semantic relationships efficiently.•TEA shows superior efficiency and effectiveness on schema matching/entity resolution.