[TDS model with transducer]

flashlight / wav2letter

Facebook AI Research's Automatic Speech Recognition Toolkit

https://github.com/facebookresearch/wav2letter/wiki

Other

6.37k stars 1.01k forks source link

[TDS model with transducer] #964

Closed jinggaizi closed 3 years ago

jinggaizi commented 3 years ago

Question

transducer loss is better than ctc in paper https://arxiv.org/pdf/2011.04785.pdf. it's use LC-BLSTM as encoder to compare 3 loss ,have you try to train a TDS-transducer, is it better than TDS-ctc?

tlikhomanenko commented 3 years ago

As far as I know we didn't try transducer, so welcome to test it too! To be sure cc @vineelpratap @xuqiantong.

Would be happy if you share later your results here.

vineelpratap commented 3 years ago

Hi @jinggaizi, we haven't tried running TDS with transducer. So, can't comment on the performance.