language	ar
tags	hassaniya mauritania dialect
task_categories	translation text-generation
license	mit

AI for RIM: Hassaniya Dialect Datasets 🇲🇷

This repository contains open-source datasets designed for training and fine-tuning Large Language Models (LLMs) on the Hassaniya dialect of Mauritania.

Datasets

The repository currently hosts two primary datasets in JSONL format, ready for LLM training:

1. Translation Dataset (ar_to_hs_translation.jsonl)

• Content: Parallel corpus of Standard Arabic to Hassaniya translation pairs.
• Size: ~4,430 pairs.
• Format: Chat-style structure suitable for instruction tuning.

  {"messages": [{"role": "user", "content": "Translate the following to Hassaniya: ..."}, {"role": "model", "content": "..."}]}

2. Customer Support Dataset (hassaniya_customer_support.jsonl)

• Content: Multi-turn customer support conversations entirely in Hassaniya.
• Size: ~594 conversation turns.
• Format: System, User, and Assistant role structure.

  {"messages": [{"role": "system", "content": "..."}, {"role": "user", "content": "..."}, {"role": "model", "content": "..."}]}

Goals

The goal of this project is to build foundational resources for Hassaniya NLP and AI, making advanced language technologies accessible to the Mauritanian community.

Upcoming

• Hassaniya Text Normalization Tool: A standardized tool for normalizing Hassaniya text is currently in development and will be released soon.

Acknowledgements

Special thanks to Dr. Ahmed Oumar for his invaluable guidance, mentorship, and support in making this project possible.