About this document

TableBank: Dataset for Table Detection by Throw Away is a document available to read on EtoBox.

This document introduces TableBank, a new dataset for table detection and recognition created using weak supervision from Word and Latex documents on the internet. TableBank contains over 417,000 high quality labeled tables in a variety of domains. Several strong baselines were developed using state-of-the-art models, which showed that layout and format variation significantly impacts table analysis accuracy. Models trained on one domain did not perform well on other domains, indicating room for improvement

Author

Throw Away

Language

EN