metadata

dataset_info:
  features:
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 5261859457
      num_examples: 47978340
    - name: test
      num_bytes: 46574356
      num_examples: 437561
  download_size: 2776698934
  dataset_size: 5308433813
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
license: odc-by
language:
  - ar
tags:
  - Fineweb2
  - Arabic
  - Datasets
  - Najdi
pretty_name: 'Najdi Arabic '
size_categories:
  - 10M<n<100M

FineWeb2 Najdi Arabic

🇸🇦 This is the Nadj Arabic Portion of The FineWeb2 Dataset.

🇸🇦 This dataset contains a comprehensive collection of text in Najdi Arabic, a regional dialect within the Afro-Asiatic language family. With over 562 million words and 1.6 million documents, it provides a valuable resource for developing NLP tools and applications specific to Najdi Arabic.

Purpose of This Repository

This repository provides easy access to the Arabic portion - Najdi of the extensive FineWeb2 dataset. My primary goal is to make this valuable data more accessible and impactful for researchers, developers, and anyone working on Arabic natural language processing (NLP) projects.

By focusing on Arabic, I aim to:

Simplify Access: Provide a direct and streamlined way to download the Arabic portion of the dataset without navigating through the larger collection.
Promote Research: Enable more efficient use of Arabic text data for NLP, LLMs, and linguistic research.
Empower the Community: Support Arabic language processing and contribute to the growth of multilingual NLP capabilities.
Encourage Collaboration: Foster an environment where researchers and developers can build impactful applications using Arabic data.

Credit to the Original Work

The dataset is released under the Open Data Commons Attribution License (ODC-By) v1.0, with additional usage subject to CommonCrawl's Terms of Use.

Citation

If you use this dataset, please cite it as follows:

@software{penedo2024fineweb-2,
  author = {Penedo, Guilherme and Kydlíček, Hynek and Sabolčec, Vinko and Messmer, Bettina and Foroutan, Negar and Jaggi, Martin and von Werra, Leandro and Wolf, Thomas},
  title = {FineWeb2: A sparkling update with 1000s of languages},
  month = dec,
  year = 2024,
  doi = {10.57967/hf/3744},
  url = {https://huggingface.co/datasets/HuggingFaceFW/fineweb-2}
}