Shield:

This work is licensed under a Creative Commons Attribution 4.0 International License.

We introduce a novel dataset called CORD, which stands for a COnsolidated Receipt Dataset for post-OCR parsing.

OCR is inevitably linked to NLP since its final output is in text. Advances in document intelligence are driving the need for a unified technology that integrates OCR with various NLP tasks, especially semantic parsing. Since OCR and semantic parsing have been studied as separate tasks so far, the datasets for each task on their own are rich, while those for the integrated post-OCR parsing tasks are relatively insufficient. In this study, we publish a consolidated dataset for receipt parsing as the first step towards post-OCR parsing tasks. The dataset consists of thousands of Indonesian receipts, which contains images and box/text annotations for OCR, and multi-level semantic labels for parsing. The proposed dataset can be used to address various OCR and parsing tasks.

• CORD v2 data has been uploaded to the Hugging Face Datasets. We investigated all data and corrected the incorrect labels. Also, we added the attribute sub_group_id to each element of valid_line. We can use this information to describe more accurate hierarchy of the resulting parse. See the gt_parse of the examples containing the menu.sub_nm, and compare those of the CORD v1. [20220720]
• CORD v1 data has been uploaded to the Hugging Face Datasets. CORD v1 has the same contents as v0 except the gt_parse attribute. gt_parse represents the parse format constructed from the valid_line. [20220720]
• 1,000 sample dataset will be available soon. Some class labels shown in the original paper were removed due to Indonesian legal issues. In particular, the store_info, payment_info, and etc fields have been removed from the target class to be published. [20191212]
• 1,000 sample dataset has been released. [20191226]
• Some categories not used in the current dataset have been removed from the class definition. [20200210]

• Large Scale: over 11,000 Indonesian receipts collected from shops and restaurants
• Fine-grained classes: five superclass and 42 subclass labels
• Multi hierarchy: includes group annotations
• Additional information: line group (row_id), region of interest (roi), cut lines (repeating_symbol), and is_key flag

*A blank 'roi' value means the entire area of the image.

@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}
}

@article{hwang2019post,
  title={Post-OCR parsing: building simple and robust parser via BIO tagging},
  author={Hwang, Wonseok and Kim, Seonghyeon and Yim, Jinyeong and Seo, Minjoon and Park, Seunghyun and Park, Sungrae and Lee, Junyeop and Lee, Bado and Lee, Hwalsuk}
  booktitle={Document Intelligence Workshop at Neural Information Processing Systems}
  year={2019}
}

@article{kim2021donut,
   title={OCR-free Document Understanding Transformer},
   author={Kim, Geewook and Hong, Teakgyu and Yim, Moonbin and Nam, JeongYeon and Park, Jinyoung and Yim, Jinyeong and Hwang, Wonseok and Yun, Sangdoo and Han, Dongyoon and Park, Seunghyun},
   journal={arXiv preprint arXiv:2111.15664},
   year={2021}
}