4 ms·
Espnet/espnet: End-To-End Speech Processing Toolkit
- gok 8y agoJust to put those WER numbers into context: the LF-MMI Kaldi paper from 2016 had 4.28%, this readme mentions 7.1%. Still, looks cool!
- gok 8y agoErg too late to edit but I meant to say this was on LibriSpeech
- peteforde 8y agoFor those of us who are not speech processing experts, what does this toolkit allow someone who invests the time to achieve? A personal pet peeve of mine is project READMEs that project an implicit "if you're reading this, you already know what it's for" voice. What behaviour or previous technology does this replace? What other packages could I compare it to? Is this exciting for a reason that other end-to-end speech processing toolkits are not? Throw us a bone.
- ianai 8y agoAgreed, sell people on it - including the noobs.
- singularity2001 8y agoHow long to get this thing running? Is it usable for end-users?