w2v-bert-2.0 Swahili (400h)
A facebook/w2v-bert-2.0 model fine-tuned for Swahili automatic speech recognition (CTC) on ~400 hours of Swahili speech.
Training data
Fine-tuned on a combined ~400h Swahili corpus from four public datasets:
- Common Voice (CV) — crowd-sourced read speech
- FLEURS — Google FLEURS read speech
- AMMI — African Masters in Machine Intelligence Swahili speech
- ALFFA — African Languages in the Field: speech Fundamentals and Automation