Research Saturday Week 2: The NaijaVoices Dataset

​This week at Research Saturday we will review NaijaVoices ; a ground-breaking open-source speech dataset that captures the richness of Nigeria’s languages and accents, including Yoruba, Igbo, Hausa, Pidgin, and Nigerian English.

​We’ll explore how Emezue, C. et al. (2025) built NaijaVoices through community-driven data collection, annotation, and cultural representation and what lessons The Datum Africa Community can draw from it.

​Together, we’ll analyse the paper’s methodology, identify gaps in current African speech data infrastructure, and practical guidelines for inclusive, ethical, and scalable African voice data collection.

​Paper: https://arxiv.org/pdf/2505.20564

​Presenter: Chiamaka Ikeh (AI Researcher, Datum Africa)