Natural Language Processing

Development and Evaluation of Large Language Models for Under-resourced Ethiopian Languages

This research proposes developing and evaluating a multilingual Large Language Model (LLM) for under-resourced Ethiopian languages. It focuses on building quality multilingual datasets, improving model training through techniques such as transfer learning and fine-tuning, and evaluating performance across tasks including translation, question answering, summarization, text generation, and sentiment analysis. The study also addresses fairness, hallucination, cultural accuracy, code-switching, dialect variation, and limited data. The expected result is an open, reproducible framework with datasets, benchmarks, methodologies, and a prototype Ethiopian-language LLM that can support inclusive AI, education, digital language preservation, and public services.

100

Progress

15

Researchers