# FUNSD: A Dataset for Form Understanding in Noisy Scanned Documents

## TL;DR

The paper introduces FUNSD, a comprehensive dataset of 199 noisy scanned forms with detailed annotations, enabling research in form understanding tasks like text detection, OCR, and layout analysis.

## Contribution

It provides the first publicly available, fully annotated dataset for form understanding in noisy scanned documents, along with baseline methods and evaluation metrics.

## Key findings

- Dataset contains 199 annotated forms with diverse noise and appearance.
- Baseline models and metrics are established for form understanding tasks.
- The dataset facilitates advancements in OCR, layout analysis, and entity linking.

## Abstract

We present a new dataset for form understanding in noisy scanned documents (FUNSD) that aims at extracting and structuring the textual content of forms. The dataset comprises 199 real, fully annotated, scanned forms. The documents are noisy and vary widely in appearance, making form understanding (FoUn) a challenging task. The proposed dataset can be used for various tasks, including text detection, optical character recognition, spatial layout analysis, and entity labeling/linking. To the best of our knowledge, this is the first publicly available dataset with comprehensive annotations to address FoUn task. We also present a set of baselines and introduce metrics to evaluate performance on the FUNSD dataset, which can be downloaded at https://guillaumejaume.github.io/FUNSD/.