Great work
Hi Stephenn ... fellow Konkani lover here... What was the size of the dataset that you trained this model on ?
Have you been working on improving this model ?
Hi @thak123 , thanks for your interest in this model. However i would not recommend using the current model as the current model was just uploaded for testing purposes and wasn't trained efficiently.
We at LDCIL (Langauge Data Consortium for Indian Languages) www.ldcil.org have around ~ 70 hrs of high quality speech dataset + additional Konkani speech datasets.
Give me a couple of weeks, i am planning to released a really effective and strong konkani ASR model trained on much better dataset and much better architecture than the current architecture.XLSR-conformer-Large with LM , further we also have plans to released a fine-tuned version of Whisper in konkani.
would keep you posted here when we launch.
Hi @StephennFernandes is the dataset released to public ?