Projects

MEY-D26: a 500-hour Dioula speech corpus Ongoing

Dioula is spoken by millions of people across Côte d’Ivoire, yet very little recorded and transcribed speech exists for it. MEY-D26 collects 500 hours of speech to train and evaluate speech recognition and speech synthesis models.

Funding

Masakhane sub-grant, led by Centre IDTDS

Period

July 2026 to April 2027

Partners

UFHB and CURAT, OIA ANACARDE

Content

150 hours of read speech and 350 hours of spontaneous speech, in five domains: health, agriculture, education, finance and public administration

Varieties

Vehicular Dioula (Tagbusi), Kong Dioula and Worodougou Dioula

Places

Abidjan, Bouaké, Korhogo, Séguéla, Bondoukou, Kong and surrounding villages

Contributors read sentences written by linguists or answer open questions in Dioula. Four Ivorian linguists check every recording and transcribe it. A subset of high-quality read speech will serve for text-to-speech. A doctoral student working with me will extend it into a thesis.

KAN, a platform for collecting language data In use

KAN is the web platform we built at Centre IDTDS to collect and validate speech in any language. Contributors choose a language, record read or spontaneous speech, and earn points. Linguists validate recordings in two stages, first the audio and then the transcription. Spontaneous speech receives a draft transcription from Meta’s MMS speech recognition model, which linguists correct instead of typing from scratch.

The platform serves MEY-D26 today and is designed to take other Ivorian languages next. Open KAN

Voix d’Ivoire Ongoing

Voix d’Ivoire is a community of more than 400 people who record, listen to and check data in Ivorian languages. It started with Dioula in October 2026 and is open to speakers of every language of the country. Its motto: “a community to bring our languages into the digital world and AI”. The community is coordinated by Sory Diomandé.

If you speak an Ivorian language and want to contribute, write to me.

Musow Kènèya, a maternal health voice assistant Pilot

Most health information in Côte d’Ivoire is written in French. Musow Kènèya (“women’s health” in Dioula) lets them ask a question about pregnancy or newborn care out loud, in Dioula, and hear the answer spoken back. The pilot is run by NOUR TECH, with Dr Awa Fofana as medical director.

Pro.IA Ongoing

Pro.IA is a five-year national programme to develop and popularise artificial intelligence, led by UNDP, UNESCO, CGECI (the employers’ federation) and Université Félix Houphouët-Boigny. I have sat on its steering committee since 2023. It includes an AI prototyping centre where companies and students can turn ideas into working prototypes. I am currently looking for national and international partners to equip it.

Centre IDTDS

I chair Centre IDTDS (Intelligent Digital Transformation and Data Strategies), a non-profit research and training centre based in Abidjan-Cocody. It hosts the MEY-D26 project, KAN and the Voix d’Ivoire community, provides the legal framework for IndabaX Côte d’Ivoire, and has been a member of Research4Life since October 2026. Visit idtdscenter.org