MTL with Kaldi RNN-LM Training

[Roshan S Sharma]

Hi,

Is there a way to add another loss function and modify the RNNLM Architecture so that the RNNLM is trained for domain classification also? And so that it produces a domain label as additional output that I can integrate into the Kaldi RNNLM Rescoring Framework? I think I can do the latter part, i.e., fusing it into the Lattice Rescoring, but am not sure about how to do the first.

How do I modify the architecture and add the Additional Loss?

Thanks

Roshan S. Sharma

[Daniel Povey]

Sorry, that wouldn't be possible without rather complex code changes -- or at least, lots of C++-level code changes :-(

--
Go to http://kaldi-asr.org/forums.html find out how to join
---
You received this message because you are subscribed to the Google Groups "kaldi-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+...@googlegroups.com.
To post to this group, send email to kaldi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kaldi-help/CAJ1Uch4jyyPUYXsS_JMYWAFK%2BTFqo6Xp-7OfToR7FuGV6W8n1g%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.