I’m happy to announce TIDIGITSRecipe an entirely Julia-flavoured automatic-speech-recognition (ASR) recipe!
The repository comes with a live demo where you can use your own voice to recognise spoken digits in English. I’d be happy if some of you could try it out and give me some feedback.
This repo is not a package but a set of scripts that can be used to train an ASR system.
The following packages are used (among others):
- Flux as ML library
- HMMGradients for maximum likelihood training
- FiniteStateTransducers for WFST compositions
As far as I know I think this is the first attempt of a completely Julia-based ASR toolkit. Of course TIDIGITS is a relatively simple ASR problem. Hopefully this can be a first-step towards more challenging and exciting datasets!