📋 Tender Overview
PROJECT ID
110732
TOTAL FUNDING
CAD 5,360,000.00
PROJECT STATUS
Active
DURATION
37 months
Summary
Artificial intelligence (AI) systems are only as robust and equitable as the data they are trained on. However, African languages remain grossly underrepresented in the foundational datasets powering modern AI. Despite accounting for over one billion speakers across more than 2,000 languages, African linguistic and cultural diversity is largely absent from large-scale language models, speech systems and generative AI platforms. This exclusion not only limits the performance and usability of AI tools in African contexts, it also deepens digital inequities, particularly for rural populations, women and speakers of indigenous or unwritten languages.
The Masakhane AI Hub for African Languages is a strategic response to this foundational gap. It will develop comprehensive, open-source, culturally grounded multimodal datasets across 40 African languages covering text, speech and visual datasets. There will be a deliberate focus on gender balance, cultural/regional diversity and bias mitigation. This will enable the development of inclusive AI tools tailored to African realities — like speech-based services in health care, multilingual education platforms and local language assistants in agriculture. By anchoring this work in Masakhane, which is a grassroots, expert-led organization that champions participatory methods, the project aims to shift Africa from being a data-poor consumer of AI technologies to a globally recognized contributor to ethical and inclusive AI innovation.
This project is a part of the Artificial Intelligence for Development program, a five-year partnership between IDRC and the United Kingdom’s Foreign, Commonwealth and Development Office to support policy, innovations and expanded leadership to spur safe and responsible AI development and use in Africa. The vision of the new partnership is to support an inclusive responsible AI ecosystem that improves quality of life for all in Africa and beyond.
