Search
Google has recently rolled out a substantial speech dataset in Africa, named WAXAL, encompassing 21 African languages, including Acholi, Hausa, Luganda, and Yoruba. These languages collectively represent over 100 million speakers. The dataset was developed through a collaborative effort involving local African academic and community institutions, and it comprises roughly 1,250 hours of automatic speech recognition (ASR) data, along with 180 hours of high-quality text-to-speech (TTS) data. Notably, the ownership of this dataset remains with African institutions. This project is designed to tackle the shortage of resources for African languages within mainstream AI systems, fostering the development of inclusive voice technology and safeguarding digital language heritage. The dataset is now freely accessible to researchers worldwide under the CC-BY-4.0 license.
-
C114 Communication Network -
Communication Home
