I try to generate a language model like described here for the German Voxforge Model.
But it does not work? Obviously, the german voxforge model uses other .dic
syntax/letters than the "normal" models like wsj/digits/hub.
First I thought there is this perl script espeak2phones.pl for exactly doing
this, but this is just not working anymore. (since espeak changed a bit). So
how to make a own language (.dic and .lm) model for german voxforge?
(The thing I want to make is a "little" dictionary only for one/two-word
commands. It works not bad with wsj model, but I think it would be better with
the german model.)
Thx!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Language model will only contain words from the first column in your
dictionary. For ex. ANALOGIE
So for creating LM you don't have to worry about symbols like @ which denote
the German phones I suppose.
I don't know if there is any tool which will give you phone decomposition of
German words. Obviously words in German will be different than English.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi there,
I try to generate a language model like described here for the German Voxforge Model.
But it does not work? Obviously, the german voxforge model uses other .dic
syntax/letters than the "normal" models like wsj/digits/hub.
First I thought there is this perl script espeak2phones.pl for exactly doing
this, but this is just not working anymore. (since espeak changed a bit). So
how to make a own language (.dic and .lm) model for german voxforge?
(The thing I want to make is a "little" dictionary only for one/two-word
commands. It works not bad with wsj model, but I think it would be better with
the german model.)
Thx!
The wiki page provides at least 3 alternatives for web service. You just need
to read it till the end.
As for two-three word commands, it's easier to write dictionary and jsgf
grammar yourself in a text editor.
I know the wiki page, but your post did not help me.
Seems that I always have to generate a Dictionary by hand for it.
Maybe you don't know what I mean, here an exmaple:
.dic Voxforge Germany:
ANALOG qq a n a l o: k
ANALOGE qq a n a l o: g @
ANALOGIE qq a n a l o: g i:
Generated .dic: (english)
BACKWARD B AE K W ER D
BROWSER B R AW Z ER
E-MAIL IY M EY L
FORWARD F AO R W ER D
So how to generate one for Voxforge German? Or transform the english one to
it?
Hi,
Language model will only contain words from the first column in your
dictionary. For ex. ANALOGIE
So for creating LM you don't have to worry about symbols like @ which denote
the German phones I suppose.
I don't know if there is any tool which will give you phone decomposition of
German words. Obviously words in German will be different than English.