Hello

I started creating a speech model for Julius using Deep Speech Mozilla corpus. The models will support DNN. They will be in version ENVR-v4.0 - meaning English language with various accents. My first results show really promising low WER even with my strong central european accent. I will publish the models in the beginning of January 2018 on SourceForge (https://sourceforge.net/projects/skrybotdomowy/). The models will be released under the same licence as Mozilla released their corpus, basically free to everyone for any use (commercial or academic). If you want to contact me personally you can write at leslaw.pawlaczyk at gmail dot com

Kind Regards

Leslaw

--- (Edited on 1/3/2018 5:07 pm [GMT-0600] by leslaw.pawlaczyk) ---

I am proud to anounce release of my first version of state of the art quality of speech models for English language with various accents for a modified Julius speech decoder

More info on:

https://sourceforge.net/projects/juliusmodels/files

https://github.com/palles77/julius

If you want to contact me please send an email to silesiaresearch at gmail dot com

Also please share this message if you think it could help anyone in their applications.

Best Regards

Leslaw

--- (Edited on 4/9/2018 5:25 pm [GMT-0500] by leslaw.pawlaczyk) ---

>first version of state of the art quality of speech models for English

>language with various accents for a modified Julius speech decoder

Well done!

--- (Edited on 4/10/2018 7:55 am [GMT-0400] by kmaclean) ---