metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: markdown
      dtype: string
  splits:
    - name: train
      num_bytes: 2708743071.071875
      num_examples: 13694
    - name: test
      num_bytes: 299648958.76012516
      num_examples: 1522
  download_size: 2941596774
  dataset_size: 3008392029.8320003
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
license: gpl-3.0
task_categories:
  - image-to-text
language:
  - ar
pretty_name: Arabic Img2MD
size_categories:
  - 10K<n<100K

Arabic Img2MD

Dataset Summary

The arabic-img2md dataset consists of 15,000 examples of PDF pages paired with their Markdown counterparts. The dataset is split into:

Train: 13,700 examples
Test: 1,520 examples

This dataset was created as part of the open-source research project Arabic Nougat to enable OCR and Markdown extraction from Arabic documents. It contains mostly Arabic text but also includes examples with English text.

Usage

The dataset was used to train two models:

arabic-base-nougat
arabic-large-nougat

These models are designed for OCR tasks and converting PDF content to Markdown in the Arabic language context.

Research Context

The dataset supports the findings of the research paper: Arabic-Nougat: Fine-Tuning Vision Transformers for Arabic OCR and Markdown Extraction.

Licensing

This dataset is released under the GPL-3.0 License, ensuring its open-source availability.

Citation

If you use this dataset, please cite the corresponding research paper:

@misc{rashad2024arabicnougatfinetuningvisiontransformers,
      title={Arabic-Nougat: Fine-Tuning Vision Transformers for Arabic OCR and Markdown Extraction}, 
      author={Mohamed Rashad},
      year={2024},
      eprint={2411.17835},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2411.17835}, 
}