Google, collaborating with research institutions across Africa, has introduced WAXAL, an extensive open-access speech dataset aimed at improving AI technologies for African languages.

The dataset features voice samples from 21 Sub-Saharan African languages, including Hausa, Yoruba, Igbo, Luganda, Swahili, and Acholi. Google states that WAXAL aims to serve more than 100 million speakers who have historically been excluded from voice-enabled technologies because of limited high-quality language resources.

Voice-activated assistants, transcription services, and other speech-driven technologies are widespread globally. Yet, Africa’s more than 2,000 languages have largely been overlooked in AI development due to scarce speech data, creating a digital gap that limits millions from accessing voice-enabled solutions in education, healthcare, and business.

To bridge this gap, WAXAL was created over a three-year period with funding from Google. The dataset includes 1,250 hours of transcribed natural speech and over 20 hours of premium studio recordings, enabling the development of lifelike synthetic voices.

“The real significance of WAXAL lies in empowering communities across Africa,” says Aisha Walcott-Bryantt, Head of Google Research Africa. “By offering a vital resource for students, researchers, and entrepreneurs to develop technology in their native languages, this dataset opens access to over 100 million people on their own terms.”

Community participation was central to the project. African universities and organizations such as Makerere University in Uganda, the University of Ghana, and Digital Umuganda in Rwanda spearheaded the data collection efforts, with guidance from Google researchers.

Unlike numerous international datasets, the data remains under the ownership of the partner institutions. This arrangement allows African researchers and students to create their own applications and tools independently, without depending on external corporations.

“For artificial intelligence to truly serve Africa, it must understand our languages and cultural contexts,” emphasizes Joyce Nakatumba-Nabende, Senior Lecturer at Makerere University. “The WAXAL dataset provides our researchers with the quality data required to develop speech technologies that accurately represent our diverse communities.”

At the University of Ghana, more than 7,000 volunteers lent their voices to the initiative. Professor Isaac Wiafe, an Associate Professor at the university, explains that this effort is fostering innovation across sectors like healthcare, education, and agriculture.

The WAXAL dataset is now open to the public, providing developers, researchers, and startups with essential speech data to create more inclusive AI solutions throughout Africa.