Form Understanding in Noisy Scanned Documents Dataset

FUNSD is a dataset for form understanding in noisy scanned documents. It was created by gauravsinghmmtxt and last updated in March 2022.

Use Cases

• Train a layout analysis model to detect and classify form fields like 'question' and 'answer' in noisy scans.
• Fine-tune a named entity recognition model to extract semantic entities from the text of scanned form fields.
• Develop a form reconstruction model that links related 'question' and 'answer' fields based on spatial and semantic context.
• Benchmark optical character recognition (OCR) and information extraction systems on noisy, real-world scanned documents.

Strengths

• Dataset is focused on a specific, challenging task of understanding noisy scanned forms.
• The dataset was last updated in March 2022, indicating a defined version.

Limitations

• The dataset size, number of documents, and specific annotation details are not provided in the input.
• The geographic and temporal coverage of the source documents is unknown.

Provenance

Source

huggingface, user gauravsinghmmtxt

Freshness

Last updated on 2022-03-23.