AI breaks language barrier, adds Yoruba, Hausa

Google and several African research institutions have launched a new project called WAXAL to help artificial intelligence understand and speak African languages such as Hausa, Luganda, Yoruba, and Acholi.

WAXAL is a large and free speech dataset created to support research and make AI technology more inclusive.

It helps close a digital gap affecting more than 100 million people who speak African languages that are often ignored by modern technology.

Many voice-based tools like virtual assistants and speech-to-text apps work well in major world languages, but most of Africa’s over 2,000 languages do not have enough quality speech data.

Because of this, millions of people cannot use technology in their own languages.

The WAXAL project was created to solve this problem. It took three years to develop and was funded by Google. The dataset includes 1,250 hours of transcribed everyday speech and more than 20 hours of studio-quality recordings that can be used to create natural-sounding artificial voices.

The Head of Google Research Africa, Aisha Walcott-Bryant, said the project is meant to empower Africans.

She explained that the dataset gives students, researchers, and entrepreneurs the tools they need to build technology in their own languages. This can lead to new learning tools, voice services, and economic opportunities across Africa.

An important part of the project is that it was created by Africans and for African communities.

Universities and organisations such as Makerere University in Uganda, the University of Ghana, and Digital Umuganda in Rwanda led the data collection, with support from Google. These African partners fully own the data, setting a fair example for future AI projects.

The dataset covers 21 African languages, including: Acholi, Akan, Dagaare, Dagbani, Dholuo, Ewe, Fante, Fulani, Hausa, Igbo, Kikuyu, Lingala, Luganda, Malagasy, Shona, Swahili, Yoruba, and others.