SRFUND: A Multi-Granularity Document Hierarchical Structure
Reconstruction Benchmark for Enhanced Form Understanding

A hierarchically structured multi-task form understanding benchmark.

Dataset Overview

A dataset for the document understanding community.

  • 1592 fully annotated forms

  • 529711 words

  • 112662 text-lines

  • 96824 semantic entities

  • 122594 relations

Img

Examples

Word to text-line merging, text-line to entity merging, entity category classification, item table localization and entity-based full-document hierarchical structure recovery.

Img