Voice technology infrastructure for artificial intelligence engine in Rwanda’s local language (Kinyarwanda)

Blaue Lichtwelle vor schwarzem Hintergrund
© Pawel Czerwinski

Country

  • Rwanda

Module

  • Disruptive Technologies

Speech is becoming an important way to interact with personal electronics: voice control is relevant for the future of human-machine interaction. However, developers, researchers, and startups around the globe working on voice-recognition technology face one problem alike:a lack of freely available voice data in their respective language to train speech-to-text engines.  

This often leads to limited access to voice enabled technology and deepens the digital divide for many people due to language barrier. Consequently, Rwandan technologies lack voice-interaction in the national language, Kinyarwanda, which makes voice enabled digital solutions only accessible to English or French speakers.  The goal of this solution is to make artificial intelligence (AI) technology accessible in native language. 

Developed in collaboration with Digital Umuganda and the Mozilla Foundation, the voice technology infrastructure for Kinyarwanda AI engine offers a solution in the form of an open dataset for voice-interaction. The solution is hosted on Mozilla’s Common Voice platform and has over 2400 hours of Kinyarwanda speech, of which over 2000 hours are validated. The solution is foundational for AI research and innovation in Rwanda. It allows for transfer learning to other Bantu languages, supports access to information and creation of automated public services for Rwandan nationals.