6 ms·
The WER on LibriSpeech clean test set is around 0.087 and the CER is 0.030. We trained on around a 5000 hours dataset which included LibriSpeech train.
by braindead_in 8y ago
The WER on LibriSpeech clean test set is around 0.087 and the CER is 0.030. We trained on around a 5000 hours dataset which included LibriSpeech train.
- gok 8y agoSo where did 95% come from?
- braindead_in 8y agoI think I made a mistake there. It should have been 92% accuracy. Can't find the edit button now.
- chadmeister 8y agoYou mean you meant to type 91.3% but accidently typed 95%? Were you using your own transaction software?
- gok 8y agoHave you considered using a more traditional HMM style recognizer? The stock Kaldi chain model recipe should get you more like 5 or 4.5% WER on LibriSpeech.
- braindead_in 8y agoWe actually started our experiments with Kaldi and even built a dataset out of our files to train on. But we found that Kaldi required a lot of data-prep and a long lead time. Our internal dataset is quite large and data prep is quite easy compared to Kaldi.