Links to talk
Scheduled Time: Sunday, May 28, 2023, 3:30 PM IST
Talk in 30 seconds
I talked about Malayalam language and what makes it hard to work compared to other high resource languages like English. Then I highlighted some of the Malayalam models build during Whisper event and why I went ahead to benchmark these models. These models where benchmarked on two datasets mainly Common Voice and Malayalam Speech Corpus. I also shared more about the Vegam Whisper model family for making whisper models faster. Also I shared about Malayalam speech to text demo Pallakku which I am developing now with Vegam whisper weights.