VoxForge
VoxForge
Hi,
I use TIMIT database and "beep-1.0" dictionary. However when I use the following command:
HDMan -A -D -T 1 -m -w wlist -n monophones1 -i -l dlog dict ../lexicon/voxforge_lexicon
I get this error:
Output dictionary dict opened
Source dictionary lexicon/voxforge_lexicon opened
ERROR [+5013] ReadString: String too long
In spite of all these modifications in beep dictionary as follows:
I replaced all tabs with single space;
I Put a '\' before every double quote (");
I Put a '\' before any dictionary entry beginning with single quote (')
I still have the same error message.
What could be the problem?
Thank you in advance;
Regards.
> "beep-1.0" dictionary.
From the Julius 3.2. manual:
4. Word dictionary files (-v)
Julius's word dictionary format is very similar to HTK's format. The difference is that the second field (Output Symbol) is not optional.
Format
It is necessary to use the same Japanese encoding system in the dictionary file and the word N-gram
language files so the appropriate entries can be matched.
Words that do not have N-gram entries are matched to the <UNK> N-gram. That N-gram probability
will be that of the total corrected probabilities of all non-N-gram words.
This is sent to the output queue as a recognition result. The value must be surrounded with brackets
[ ]. If the symbol is [] then nothing will be outputted.
The HMM phoneme sequence is described using monophones. (When using triphones HMM's
intraword context dependencies are automatically created when loading the dictionary)
[Example]
(It is not necessary for words to be sorted alphabetically.)
ABANDONMENT [ABANDONMENT] xb@ndInmInt
ABBAS [ABBAS] @bxs
ABBAS [ABBAS] @bxz
ABBEY [ABBEY] @bi
ABBOTT [ABBOTT] @bxt
ABBOUND [ABBOUND] xbud
ABIDE [ABIDE] xbYd
ABILITIES [ABILITIES] xbIlItiz
ABILITY [ABILITY] xbIlIti
ABLAZE [ABLAZE] xblez