Can we finetune OpenAI whisper using this dataset
I was looking for good dataset for shona for a long time, as I stumble upon this , really thankful. As I was looking in to the real time transcription using whisper, Is it valuable to use this data for whisper finetune
Yes, absolutely
We trained a CTC speech recognition model and the lowest WER we obtained was 23%
the model is here https://huggingface.co/badrex/w2v-bert-2.0-shona-asr
Do you have your own evaluation data?
I was looking for shona datasets and models , I am a native shona speaker and your resources have been helpful lately, if you have more can I please have the refs?
We have trained one speech-to-text model on the Shona dataset
https://huggingface.co/badrex/w2v-bert-2.0-shona-asr
I don't have any other datasets for Shona at the moment