Datasets:
wkrl
/

Sub-tasks:
parsing
Languages:
English
Multilinguality:
monolingual
Size Categories:
1K<n<10K
Language Creators:
crowdsourced
Annotations Creators:
crowdsourced
Source Datasets:
original
License:
cord / README.md
wkrl's picture
Update README.md
609880c
metadata
annotations_creators:
  - crowdsourced
language_creators:
  - crowdsourced
language:
  - en
multilinguality:
  - monolingual
license:
  - cc-by-4.0
pretty_name: CORD
size_categories:
  - 1K<n<10K
source_datasets:
  - original
task_categories:
  - token-classification
task_ids:
  - parsing

Dataset Card for CORD (Consolidated Receipt Dataset)

Table of Contents

Dataset Description

Dataset Summary

[More Information Needed]

Supported Tasks and Leaderboards

[More Information Needed]

Dataset Structure

Data Instances

[More Information Needed]

Data Fields

{
  "id": datasets.Value("string"),
  "words": datasets.Sequence(datasets.Value("string")),
  "bboxes": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
  "labels": datasets.Sequence(datasets.features.ClassLabel(names=_LABELS)),
  "images": datasets.features.Image(),
}

Data Splits

  • train (800 rows)
  • validation (100 rows)
  • test (100 rows)

Dataset Creation

Licensing Information

Creative Commons Attribution 4.0 International License

Citation Information

@article{park2019cord,
  title={CORD: A Consolidated Receipt Dataset for Post-OCR Parsing},
  author={Park, Seunghyun and Shin, Seung and Lee, Bado and Lee, Junyeop and Surh, Jaeheung and Seo, Minjoon and Lee, Hwalsuk}
  booktitle={Document Intelligence Workshop at Neural Information Processing Systems}
  year={2019}
}

Contributions

Thanks to @clovaai for adding this dataset.