Question Answering Research in African Languages with Large Language Models

Native African languages are grossly underrepresented in prevailing large language models (LLMs). This results in technological marginalization of native speakers, and limits their opportunity to interact with the LLMs in formal and informal settings, thereby creating a digital inequity in the global NLP landscape. However, a key contributor to this challenge is the paucity of structured data in the numerous native African languages. Democratizing access to digital knowledge bases in native languages helps to preserve linguistic and cultural heritage, while also providing safety and commercial benefits. To foster digital inclusion and bridge the knowledge gap in LLMs especially in the question answering domain; benchmark datasets, finetuned models and evaluation strategies have been proposed in recent times. This comparison provides an overview of the few research efforts in this direction currently.