Re: mlf files & the alphadigits prototype tutorial

[robert b <http://dummy.us.eu.org/robert>]

It turns out that the problem is the fs_phones file -- there are more phones in the sample
alphadigits tar ball (or zip ball in this case :-) -- format had nothing to do with it.  fs_phones
is derived from fs_models (and fs_ci_models) which is derived from fs_num_states.

So, is there a way to derive the fs_num_states and fs_ci_models files from, say, the alphadigits
trans files?

--- robert b <http://dummy.us.eu.org/robert> wrote:
> My main confusion is with my mlf files.  I don't understand how to derive the .lab files in the
> alpha_digits stuff.  There are .txt files which seem to correspond with with .lab files, but
> when
> I try to create my own mlf file, hmm_train crashes.  Here is what part of my mlf file looks
> like:
> 
> #!MLF!#
> "*/AD-1.p19.txt"
> b v y b oh b
> .
> "*/AD-1.p20.txt"
> zero eight e u r b
> .
> "*/AD-1.p21.txt"
> q m b p two j
> .
> 
> and continues.  (The phones are the actual contents of the .txt file.)  Is this correct?  Or is
> there a more strict format the mlf file?