4

I am trying to do speech recognition using RNN. It would be really helpful if I could get some suggestions on best Theano-based libraries that I can use for RNN-based speech recognition. (I am hoping to use CTC Connectionist Temporal Classification as well.)

I heard about Keras. But I would like to know whether it would be a good selection to use it before I jump into it.

2

There are no plain Theano setups just because application of RNN to speech is not trivial, you need to have a good estimation before training to make the whole system converge.

CTC implementation for speech is usually based on Kaldi speech recognition toolkit which is used to setup the whole thing with speech specifics. You can find details here:

Theano-based extension:

https://github.com/yajiemiao/pdnn

Eesen - end-to-end recognition with CTC:

https://github.com/srvk/eesen

Another implementation from Stanford:

https://github.com/amaas/stanford-ctc

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.