>i have a big problem with the special caracter like à , é , è etc

the phonemes in your pronunciation dictionary need to be in ASCII format, but your words can be unicode.

>also i have an error message when  i want to compile my dictionnary : your

>dictionnary is malformated

need more info

when i try to execute hdman with my french phoneticqlly dictionnarry .

a                    [a]                 a sp
abreuvait       [abreuvait]     a b R J v E sp
abreuvait       [abreuvait]     a b R J v E t sp
abîme            [abîme]         a b i m sp  
achats           [achats]        a S a sp
achats           [achats]        a S a z sp
acte              [acte]          a k t sp
acte             [acte]          a k t @ sp
ai                 [ai]            E sp
aime            [aime]          E m sp
ainsi           [ainsi]         U~ s i sp
applicable      [applicable]    a p l i k a b sp
applicable      [applicable]    a p l i k a b l sp
applicable      [applicable]    a p l i k a b l @ sp
arrive          [arrive]        a R i v sp
arrive          [arrive]        a R i v @ sp
attend          [attend]        a t a~ sp
attend          [attend]        a t a~ t sp
étant           [étant]            e t a~ t sp

the result of hdman is Dictionary dictr created - 11 words processed, 1 missing (étant)

so my problem is how to make htk recognize the accentued word like étant

>so my problem is how to make htk recognize the accentued word like étant

this should work - Julius recognizes Japanese words using acoustic models trained with HTK... check the HTK email archives, there should be something ther.

or maybe ask on the Julius forum.

I had a similar problem and I found out the cause was different sort behaviors. I was using linux sort command for sorting my voxforge_lexicon. The problem was gone when I used the perlsort.pl script bundled in scripts.tgz.