Logo Lanfrica

mwimlmauritania-byte/Open-Data-Mauritania

Domaine:

natural language processing

Type de record:

dataset
Créateur:
mwi
Hôte:
Open NLP datasets for Mauritania's national languages (Pulaar, Soninke, Hassaniya Arabic, Wolof) — by MWiML # 🇲🇷 Open Data Mauritania — NLP Hub > **Initiated by MWiML — Mauritanian Women in Machine Learning** > Open, community-driven NLP resources for Mauritania's national languages. --- ## 🌍 Why This Project Exists Most AI systems today perform well only in high-resource languages like English, French, or Modern Standard Arabic. **Mauritania's national languages — Pulaar, Soninke, Wolof, and Hassaniya — are severely underrepresented.** This means healthcare AI, automated public services, and language tools simply don't work for most Mauritanians. This repository is our open answer: a community hub for collecting, cleaning, and publishing NLP datasets so that anyone can build tools that serve every citizen. --- ## 📦 Datasets in This Repository ### Pulaar / Fulfulde | File | Description | Size | |---|---|---| | `datasets/pulaar/pulaar_mauritania_sentences.json` | Sentences about Mauritanian geography, institutions, culture | 81 sentences | | `datasets/pulaar/pulaar_french_parallel_corpus.json` | Pulaar ↔ French parallel translation pairs | 991 pairs | | `datasets/pulaar/pulaar_arabic_parallel_corpus_raw.json` | Pulaar ↔ Arabic parallel corpus (needs cleaning) | ~350 pairs | **Schema** (`pulaar_mauritania_sentences.json`): ```json { "sentence": "Moritani ina jeyaa e leyɗe ɓurɗe alɗude..." } ``` **Schema** (`pulaar_french_parallel_corpus.json`): ```json { "Original": "...", "Translated": "..." } ``` --- ### Soninke | File | Description | Size | |---|---|---| | `datasets/soninke/soninke_french_dictionary.json` | Soninke ↔ French bilingual dictionary | 708 entries | **Schema** (`soninke_french_dictionary.json`): ```json { "sonike": "...", "farancais": "..." } ``` --- ### Hassaniya Arabic | File | Description | Size | |---|---|---| | `datasets/hassaniya/hassaniya_stories_collection.json` | Articles and stories in Hassaniya Arabic | 100 entries | **Schema**: ```json { "title": "...", "url": "...", "category": "...", "original_language": "hassaniya", "cont …