license: other
license_name: hackathon-only
license_link: LICENSE
dataset_info:
features:
- name: text
dtype: string
- name: is_darija
dtype: int64
splits:
- name: train
num_bytes: 172722
num_examples: 1500
download_size: 88297
dataset_size: 172722
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
task_categories:
- text-classification
language:
- ar
tags:
- darija
- classification
- msa
size_categories:
- 1K<n<10K
Dataset Card for Sawalni-AI/Darija-Arabic-Classification
This dataset provides a starting point to develop a classifier for Moroccan Darija in Arabic writing.
Dataset Details
This dataset contains 1.5k classified records, 85% in Moroccan Darija and 15% in Modern Standard Arabic. Classification has been performed using internal tools by the Sawalni AI project, and is based on the Goud/Goud-sum dataset.
- Curated by: Omar Kamali
- Funded by: Omar Kamali
- Shared by: Omar Kamali
- Language(s) (NLP): Moroccan Darija, Modern Standard Arabic
- License: Exclusively licensed for use during the ThinkAI.ma 2024 hackathon.
Uses
This dataset is intended to develop classification tooling or approaches to distinguish between Moroccan Darija and MSA (Modern Standard Arabic).
Citation Information
You can cite this work as follows:
@dataset{sawalni-darija-arabic-classification,
author={ Omar Kamali },
title={ Darija Arabic Classification },
year={ 2024 },
month={ may },
url={ https://huggingface.co/datasets/sawalni-ai/darija-arabic-classification },
doi={ 10.57967/hf/2240 },
publisher={ Hugging face }
}
References
Thanks to the contributors of the Goud/Goud-sum dataset for making their work available.
@inproceedings{issam2022goudma,
title={Goud.ma: a News Article Dataset for Summarization in Moroccan Darija},
author={Abderrahmane Issam and Khalil Mrini},
booktitle={3rd Workshop on African Natural Language Processing},
year={2022},
url={https://openreview.net/forum?id=BMVq5MELb9}
}
Dataset Card Contact
Contact the Sawalni AI team here.
Learn more at
License
This dataset is intended and licensed for exclusive use during the ThinkAI.ma hackathon, edition 2024, except with written permission from the authors. Derivative models are allowed with attribution and citation, provided the models do not include significant portions of the dataset verbatim. Derivative datasets are not allowed. Given sufficient demand a public version can be made available after the hackathon upon request.