Google has teamed up with several African research institutions to launch WAXAL, a large-scale open-source speech dataset designed to support the development of artificial intelligence technologies that understand and generate African languages. The dataset includes more than 11,000 hours of speech recordings covering 21 Sub-Saharan languages, such as Hausa, Yoruba, Luganda, Swahili, Acholi, and Igbo, addressing a long-standing lack of high-quality linguistic data for the region.
WAXAL aims to close a major digital divide by giving developers, researchers, and entrepreneurs the foundational speech resources needed to build voice recognition, text-to-speech, and other AI-powered tools tailored to African languages. The initiative was developed over three years in collaboration with institutions including Makerere University (Uganda), the University of Ghana, Digital Umuganda (Rwanda), and the African Institute for Mathematical Sciences (AIMS), ensuring that local partners play a central role in data collection and governance.
A key aspect of the project is that partner institutions retain ownership of the data they gathered while making it freely available under a permissive open license. This model is intended to foster digital sovereignty — empowering African innovators to build AI systems on their own terms without dependency on external tech giants. Advocates say that having language resources rooted in local contexts will help make AI technologies more inclusive, relevant, and culturally appropriate for millions of speakers previously excluded from voice-enabled services.
The launch of WAXAL comes amid broader efforts across Africa to localize AI development and build capacity in language technologies. With thousands of languages on the continent historically underrepresented in NLP datasets, initiatives like WAXAL are seen as crucial steps toward bridging linguistic barriers, supporting education and service delivery in native tongues, and stimulating homegrown AI innovation tailored to African social and economic needs.
