Kaldi Book

[marzi...@gmail.com]

Hi Kaldi Group,

Do you have a plan or schedule to publish a detailed book about Kaldi? I mean a tutorial on theory and implementation of Kaldi toolkit (a book like HTK Book by Daniel Povey and other authors).

Regards,

Marzi.

[Olumide]

I too would *love* to see a Kaldi book!

[Daniel Povey]

Sorry, not in the short term, for time reasons... the field is moving
too fast and there are too many things we need to implement in Kaldi.
Dan

> --
> 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.
> For more options, visit https://groups.google.com/d/optout.

[marzi...@gmail.com]

Thanks Dear Dan. Would you please suggest some text book and paper that is a good start point to learn ASR systems theory and implementation? (When you have a free moment)
I think you have valuable suggestions and experince for us.

Thanks.

[Daniel Povey]

Someone else might have better suggestions than me, I have not read
any of the existing textbooks (except from parts of the HTK Book).

Dan

[marzi...@gmail.com]

very interesting :)
I am curious about the source of your perfect experience and knowledge about ASR.

[Danijel Korzinek]

Maybe we can make a short list of peoples favorite books. I'll start:

1. https://mitpress.mit.edu/books/statistical-methods-speech-recognition

2. https://web.stanford.edu/~jurafsky/slp3/

3. https://www.amazon.com/Spoken-Dialogue-Computers-Processing-Applications/dp/0122090551

[Olumide]

As Kaldi is based on WFSTs this should be useful
https://www.amazon.co.uk/Recognition-Algorithms-Finite-State-Transducers-Processing/dp/1608454738

> --
> 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

> <mailto:kaldi-help+...@googlegroups.com>.

[Behnam Asefisaray]

According to Dan's advice, Good intro to HMM based Speech Recognition:

https://www.google.com.tr/url?sa=t&source=web&rct=j&url=http://mi.eng.cam.ac.uk/~mjfg/mjfg_NOW.pdf&ved=0ahUKEwiw44b8pI_RAhVMJsAKHTPSAygQFggZMAA&usg=AFQjCNEqsr3N0QV7QXXzUatPGwMgvGEgrw&sig2=iZGEYSG9gs0uTvdJR7ibng

[Danijel Korzinek]

If we're citing online sources:

WFST bible: http://repository.upenn.edu/cgi/viewcontent.cgi?article=1010&;context=cis_papers

Dan's thesis ;-) - http://ssli.ee.washington.edu/~mhwang/pub/others/povey.pdf (but seriously, very nice read and explains a lot of decisions behind the design of Kaldi)