---
license: apache-2.0
dataset_info:
- config_name: default
  features:
  - name: src
    dtype: string
  - name: tgt
    dtype: string
  - name: bleu
    dtype: float64
  - name: tgt_index
    sequence: int64
  - name: original_filename
    dtype: string
  - name: filename
    dtype: string
  - name: page_idx
    dtype: int64
  - name: src_layout
    sequence:
      sequence: int64
  - name: tgt_layout
    sequence:
      sequence: int64
  splits:
  - name: dev
    num_bytes: 916523238
    num_examples: 49994
  - name: test
    num_bytes: 919351137
    num_examples: 50009
  - name: train
    num_bytes: 7346736959
    num_examples: 399997
  download_size: 1754136052
  dataset_size: 9182611334
- config_name: dev
  features:
  - name: src
    dtype: string
  - name: tgt
    dtype: string
  - name: bleu
    dtype: float64
  - name: tgt_index
    sequence: int64
  - name: original_filename
    dtype: string
  - name: filename
    dtype: string
  - name: page_idx
    dtype: int64
  - name: src_layout
    sequence:
      sequence: int64
  - name: tgt_layout
    sequence:
      sequence: int64
  splits:
  - name: train
    num_bytes: 916523238
    num_examples: 49994
  - name: dev
    num_bytes: 916523238
    num_examples: 49994
  download_size: 360926516
  dataset_size: 1833046476
- config_name: test
  features:
  - name: src
    dtype: string
  - name: tgt
    dtype: string
  - name: bleu
    dtype: float64
  - name: tgt_index
    sequence: int64
  - name: original_filename
    dtype: string
  - name: filename
    dtype: string
  - name: page_idx
    dtype: int64
  - name: src_layout
    sequence:
      sequence: int64
  - name: tgt_layout
    sequence:
      sequence: int64
  splits:
  - name: test
    num_bytes: 919351137
    num_examples: 50009
  download_size: 181145705
  dataset_size: 919351137
configs:
- config_name: default
  data_files:
  - split: dev
    path: data/dev-*
  - split: test
    path: data/test-*
  - split: train
    path: data/train-*
- config_name: dev
  data_files:
  - split: train
    path: dev/train-*
  - split: dev
    path: dev/dev-*
- config_name: test
  data_files:
  - split: test
    path: test/test-*
---


# ReadingBank

ReadingBank is a benchmark dataset for reading order detection built with weak supervision from WORD documents, which contains 500K document images with a wide range of document types as well as the corresponding reading order information.

Our paper "[LayoutReader: Pre-training of Text and Layout for Reading Order Detection](https://arxiv.org/pdf/2108.11591)" has been accepted by EMNLP 2021.

Refer to the official repo for more details: https://github.com/doc-analysis/ReadingBank