Empowering West African Voices: Introducing WA-VOX
West Africa, home to hundreds of millions across vibrant linguistic diversity, remains sidelined by speech AI that prioritizes English and Mandarin. WA-VOX changes that by assembling a robust, 850-hour corpus of scripted and unscripted speech capturing everyday conversations, market chatter, and formal discussions in languages like Hausa, Yoruba, Zulu etc. This isn't just data, it's a call to action for equitable tech, built through open collaborations with local universities and communities in Nigeria, Ghana, Senegal, Mali, and Burkina Faso.
Why It Matters
Current datasets like Common Voice offer mere scraps (under 20 hours for Hausa), limiting real world ASR tools and perpetuating access barriers. WA-VOX fills this void with deep domain coverage (agriculture to culture), dialectal variety, and rigorous quality controls. As an independent researcher, I've designed it for sustainability: free for non commercial use, with a Creative Commons license and ongoing community stewardship.
Next Steps and Access
This document outlines the full roadmap from recruitment of 1,200 diverse speakers to risk mitigated collection and open release. Download to explore methodologies, baselines, and ethics protocols. Join the movement: collaborate, cite, or fund to power the next wave of inclusive language tech. Questions? Reach out via royalrubee@gmail.com.