Retrieval-based Transformer for Table Augmentation

Abstract

Data preparation, also called data wrangling, is considered one of the most expensive and time-consuming steps when performing analytics or building machine learning models. In this paper, we introduce a retrieval augmented transformer model that is self-trained for table augmentation tasks such as row and column population and data imputation. The model is trained by randomly ablating tables from the corpus and reconstructing the partial tables given as input. It substantially outperforms supervised statistical methods and transformer-based models on EntiTables and WebTables.