IbrahimAmin/arz-en-parallel-corpus

Egyptian Arabic - English Parallel Corpus 🇪🇬✨🇬🇧

Dataset Description

This dataset is a cleaned and filtered merge of multiple Egyptian Arabic - English parallel corpora, containing ~27,000 aligned sentence pairs. It’s designed for researchers and developers working on machine translation, speech translation, and other NLP tasks involving Egyptian Arabic and English.

---

Sources 📚

This dataset integrates and refines data from the following publicly available sources:

• ArzEn-MultiGenre Dataset by Hesham Haroon
• Egyptian_English_parallel Dataset by Hesham Haroon
• ArzEn_MultiGenre_subtitles Dataset by arbml
• ArzEn-ST Corpus from ArzEn Corpus Resources

---

Languages 🗣️

• Egyptian Arabic (arz)
• English (en)

---

Size 📏

~27,000 parallel sentence pairs

Split	Number of Samples
Train	25,000
Test	1,851
Total	26,851

---

Applications 🚀

Ideal for:

• Machine Translation (MT)
• Speech Recognition (ASR) with translation support

---

Preprocessing 🧹

• Removed rows with only numbers or non-informative tokens.
• Removed special tags like [HES], [LAUGHTER], and any content inside brackets (e.g., [tag]) while avoiding extra spacing.
• Filtered out missing or empty translations.
• Eliminated duplicate samples based on full row content.
• Shuffled the dataset to randomize sample order.

---

Usage Example

from datasets import load_dataset

dataset = load_dataset("IbrahimAmin/arz-en-parallel-corpus")
print(dataset['train'][0])

---

License 📄

MIT License — freely usable for research and commercial purposes

---

👏 Acknowledgement

Special thanks to Hesham Haroon, the arbml team and Injy Hamed for creating and sharing these valuable datasets.

---

Citation 📝

If you use this dataset, please cite:

@inproceedings{hamed-etal-2022-arzen,
  title     = {ArzEn-ST: A Three-way Speech Translation Corpus for Code-Switched Egyptian Arabic-English},
  author    = {Hamed, Injy and Habash, Nizar and Abdennadher, Slim and Vu, Ngoc Thang},
  booktitle = {Proceedings of the Seventh Arabic Natural Language Processing Workshop (WANLP)},
  pages     = {119--130},
  year      = {2022},
  address   = {Abu Dhabi, United Arab Emirates (Hybrid)},
  publisher = {Association for Computational Linguistics},
  url       = {https://aclanthology.org/2022.wanlp-1.12},
  doi       = {10.18653/v1/2022.wanlp-1.12}
}

@article{al-sabbagh-2024-arzen-multigenre,
  title     = {ArzEn-MultiGenre: An aligned parallel dataset of Egyptian Arabic song lyrics, novels, and subtitles, with English translations},
  author    = {Al-Sabbagh, Rania},
  journal   = {Data in Brief},
  volume    = {54},
  pages     = {110271},
  year      = {2024},
  publisher = {Elsevier},
  doi       = {10.1016/j.dib.2024.110271},
  url       = {https://www.sciencedirect.com/science/article/pii/S2352340924002403}
}

@misc{amin2025arzenparallel,
  author       = {Amin, Ibrahim},
  title        = {Egyptian Arabic - English Parallel Corpus},
  year         = {2025},
  url          = {https://huggingface.co/datasets/IbrahimAmin/arz-en-parallel-corpus},
  note         = {MIT License. Curated and cleaned from multiple public datasets.}
}