metadata
license: mit
task_categories:
- text-generation
language:
- lg
pretty_name: fine luganda
size_categories:
- 100K<n<1M
Luganda Pretraining Dataset
📝 Dataset Summary
The Luganda Pretraining Dataset is a continuously growing corpus of text in Luganda, one of Uganda’s major Bantu languages. Curated by SAGE POND, this dataset was designed to power language model pretraining, text generation, and other Luganda NLP research tasks.
Given Luganda’s low-resource status in the NLP landscape, this dataset aims to capture the linguistic richness, cultural expressions, and dialectal diversity of the language. It encompasses multiple domains including oral literature, social media, news, religious texts, and academic materials.
This dataset will be continuously updated to include:
- Dialectal and regional variations
- Contemporary colloquialisms and emerging terms
- Proverbs, idioms, and culturally significant expressions
- Domain-specific corpora (legal, health, finance, etc.)
It forms part of SAGE POND’s mission to make AI accessible to everyone, with a special focus on underrepresented languages and communities.
📚 Supported Tasks
- Causal Language Modeling (CLM)
- Masked Language Modeling (MLM)
- Text Generation
- Luganda-centric NLP research
🌍 Languages
lug— Luganda