Since MMS is unable to accurately capture numerals, the following is the structure of the vocab.json file. This format was derived from a 100-hour Akan audio dataset used during our ASR and TTS training.

Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment