Google AI Releases WAXAL: A Multilingual African Speech Dataset for Training Automatic Speech Recognition and Text-to-Speech Models

WAXAL was released as an open multilingual speech dataset, google and collaborators published WAXAL, an open multilingual speech dataset covering 24 African languages.

Covers 24 languages with ASR and TTS components, wAXAL covers 24 African languages and includes separate ASR and TTS components.

ASR component was built from transcribed natural speech collected using image-prompted prompts, the ASR component was created from transcribed natural speech collected via image-prompted descriptions.

ASR recordings were captured in natural environments with at least 15 seconds per recording, recordings for ASR were captured in speakers' natural environments with a minimum duration of 15 seconds per recording.

ASR transcriptions comprise approximately 10% of the recorded ASR audio, about 10% of the total recorded ASR audio was transcribed by paid local linguistic experts.