Digital Umuganda — Building Open Voice Data and AI Services for Kinyarwanda, Rwanda's National Language

Digital Umuganda grew out of a 2019 Kigali hackathon organised with Mozilla, applying the Rwandan tradition of communal work ("umuganda") to a digital problem: Kinyarwanda, spoken by over 12 million people, had almost no voice data on major speech-AI platforms. Working through Mozilla's open-source Common Voice project and funded by the German development agency's "FAIR Forward — AI for All" programme (BMZ/GIZ), the team organised community recording drives; by May 2021 the Kinyarwanda dataset had grown to more than 1,700 hours of speech from over 840 contributors, becoming one of Common Voice's largest non-English datasets.

Building on that data, Digital Umuganda developed "Mbaza," a Kinyarwanda voice-and-text chatbot created with GIZ to deliver COVID-19 health information to citizens who could not read English or French, and has continued developing Kinyarwanda speech-to-text infrastructure since, with active technical releases through 2025 under CEO Audace Niyonkuru. Mozilla has published a dedicated account of the design lessons from the project — including the difficulty of standardising Kinyarwanda orthography for transcription and recruiting speaker samples balanced by age, gender and dialect — a useful documented-learning record, though most reach figures are self-reported by the partner organisations rather than independently audited.

Read the full analysis: https://www.mozillafoundation.org/en/blog/lessons-from-building-for-kinyarwanda-on-common-voice/