Hello,

I have a new user of Voxforge - HTK -Julius.

I try to make a french model but I bolck on step 8 at this line

HVite -A -D -T 1 -l '*' -o SWT -b SENT-END -C config -H hmm7/macros -H hmm7/hmmdefs -i aligned.mlf -m -t 250.0 150.0 1000.0 -y lab -a -I words.mlf -S train.scp dict monophones1> HVite_log

and I have this error :

ERROR [+8220] LatticeFromLabels: Word t not defined in dictionary
FATAL ERROR - Terminating program HVite

the content of my hmm7 directory is

hmm7/hmmdefs : https://dl.dropbox.com/u/93726840/hmmdefs

hmm7/macros : https://dl.dropbox.com/u/93726840/macros

dict : https://dl.dropbox.com/u/93726840/dict

prompts : https://dl.dropbox.com/u/93726840/prompts

lexicon : https://dl.dropbox.com/u/93726840/lexicon

wlist : https://dl.dropbox.com/u/93726840/wlist

 words.mlf : https://dl.dropbox.com/u/93726840/words.mlf

train.scp : https://dl.dropbox.com/u/93726840/train.scp

and files *.lab and *.mfc : https://dl.dropbox.com/u/93726840/interim_files.zip

Can you help me to resolve this problem?

I have search on the net but I have see nothing solution

thanks a lot

You just need to read the message

>ERROR [+8220] LatticeFromLabels: Word t not defined in dictionary

It says that word "T" is not defined in a dictionary. Your dictionary is called "dict". If you open this dictionary file in a text editor you will see there is no line with the word "T". It is a problem you are looking for. You need to add a line with the word "T" to the dictionary like it's present in lexicon file already.

Try to read the messages software gives you.

hello thanks for replie,

I have already try to add the words T in dict as this :

T  [T] t t t t

but same error ...

in the dic file I have some word start with letter T in the word

TA              [TA]            t t a sp

TALONS          [TALONS]        t t a l o on s s sp

TEL             [TEL]           t t e e el sp

TENTA           [TENTA]         t t e e en t a sp

TENTACULES      [TENTACULES]    t t e e en t a c u l e e es sp

TERRE           [TERRE]         t t e e e e e r rr e e sp

TOUS            [TOUS]          t t ou o s s s sp

TOUT            [TOUT]          t'u sp

TOUTE           [TOUTE]         t t ou o t e e sp

TOUTES          [TOUTES]        t t ou o t e e es sp

TRAITE          [TRAITE]        t t r a ai ai t e e sp

TRAVAIL         [TRAVAIL]       t t r a v a a ai ai il i sp

TU              [TU]            t t u sp

it why I don't understand. 

this error means : Less the LETTER T or less a word with the letter T ?

another question for construc phoneme I use the command :

espeak -v fr -q -X "T"      

Translate 't'

 22     t (_     []

  1     t        [t]

  5     _) t     [t]

 26     _) t (_  [te]

 t'e

but I m not sure I take the good phomene 

for this example what is the phomene ?

me I have take the collum 2 so t t t t .

it's correct ?

Or what other tools I can use to construct phomene (in command line if possible)

thanks a lot

Ok, after some more investigation the problem is different. The thing is that label names are case sensitive. So as it given in the train.scp it tries to load labels with lowercase:

mfcc/sample1.mfc

mfcc/sample2.mfc

mfcc/sample3.mfc

mfcc/sample4.mfc

mfcc/sample5.mfc

However, your words.mlf has uppercase labels for utterances:

#!MLF!#

"*/SAMPLE1.lab"

TOUTE

CETTE

So it fallbacks to the lab files which have phonemes like t and i and not the words. That why it fails to proceed and reports there is no word t in the dictionary.

You need to make sure that all the labels are in lowercase. The easiest way would be to fix words.mlf file to contain lowercase labels:

#!MLF!#

"*/sample1.lab"

TOUTE

CETTE

For the reference check tutorial example

http://www.voxforge.org/uploads/Ni/jv/NijveUP5RI5o3gSCeoRlmg/words.mlf

hello,

i have do the modification in the file words.mlf

SAMPLE --> sample

and this step it's without error but the nexte step :

HERest -A -D -T 1 -C config -I aligned.mlf -t 250.0 150.0 3000.0 -S train.scp -H hmm7/macros -H hmm7/hmmdefs -M hmm8 monophones1 

HERest -A -D -T 1 -C config -I aligned.mlf -t 250.0 150.0 3000.0 -S train.scp -H hmm7/macros -H hmm7/hmmdefs -M hmm8 monophones1 

HTK Configuration Parameters[10]

  Module/Tool     Parameter                  Value

#                 NUMCEPS                       12

#                 CEPLIFTER                     22

#                 NUMCHANS                      26

#                 PREEMCOEF               0.970000

#                 USEHAMMING                  TRUE

#                 WINDOWSIZE         250000.000000

#                 SAVEWITHCRC                 TRUE

#                 SAVECOMPRESSED              TRUE

#                 TARGETRATE         100000.000000

#                 TARGETKIND          MFCC_0_D_N_Z

HERest  ML Updating: Transitions Means Variances 

 System is SHARED

99 Logical/99 Physical Models Loaded, VecSize=25

2 MMF input files

Pruning-On[250.0 150.0 3000.0]

 Processing Data: sample1.mfc; Label sample1.lab

 Utterance prob per frame = -5.294599e+01

 Processing Data: sample2.mfc; Label sample2.lab

  ERROR [+7332]  CreateInsts: Cannot have Tee models at start or end of transcription

thanks

>  ERROR [+7332]  CreateInsts: Cannot have Tee models at start or end of transcription

Tee model is the model named "sp". You have "sp" in the end. The reason for this is that in your lexicon SENT-START and SENT-END are not transcribed as sil but as something different. Fix your lexicon and run the training from initialization step

SENT-END        []              sil
SENT-START      []              sil