tchamna/LLM-Under-Ressourced-Languages

AI's progress relies on data, but African languages lack resources for development. Fine-tuning models suits Western languages, not tonal African ones. Tonal variations impact word meanings, demanding accurate representation. Addressing these nuances is crucial for AI advancements in African languages. ### Advancing AI for African Languages AI development is data-driven, yet African languages often lack the necessary resources. Current fine-tuning approaches are tailored predominantly for Western languages, which do not account for the subtleties of tonal African languages. In these languages, tonal variations are crucial as they can alter meanings, necessitating models that can accurately capture these details. This is vital for pushing AI forward in the context of African linguistics. #### Building a Language Model for Nufi In this project, we embarked on constructing a Language Model (LM) for the Nufi language—a language with scarce data resources. Our initial phase focused on developing a predictive model capable of anticipating the subsequent word in a sequence. Following this, we progressed to generating a model that could create coherent sentences not found in the existing corpus, demonstrating an understanding of language nuances beyond mere data replication.