There’s recently Common Voice from Mozilla, which is a huge free English dataset (1500 hours and growing), and wer_are_we [1] has shown really impressive accura

There’s recently Common Voice from Mozilla, which is a huge free English dataset (1500 hours and growing), and wer_are_we [1] has shown really impressive accuracy increases in published research the past few years. Exciting times. [1] https://github.com/syhw/wer_are_we https://github.com/syhw/wer_are_we