VoxForge
VoxForge
Don't have a tree.hed so can't go further to define a tiedlist. So the normal way doesn't work for me.
I succeded in running my own AM in live mode with julius and in batch mode with 50 recording and got the juliusOutput.
I'd like to know what results my model can tell me from my juliuOutput. Only the cmscore1 Line, like the following
sentence1: <s> NOVANTUNO EURO DICIOTTO </s>
wseq1: 0 12 5 10 1
phseq1: sil | n o v a n t u1 n o | E1 u r o | d i tS O1 t t o | sil
cmscore1: 1.000 0.999 0.951 1.000 1.000
tells me word recongition probabilities for each sentence.
Are there better ways to get compact result (as with HResults) also for monophones. Does anyone know how to?
thanks all.
Occimanete
Hi Occimanete,
>Don't have a tree.hed so can't go further to define a tiedlist. So the
>normal way doesn't work for me.
I believe SphinxTrain can create one automatically, though it is not as good as one that is created manually.
From my post in this thread:
For more information on how to create a tree.hed file for a new language, see the following links:
See also the HTK manual.
>Are there better ways to get compact result (as with HResults) also for
>monophones.
I created a script to convert Julius results to HTk format, so that is can be used with HResults, see here: Test Acoustic Model Using Julius
Ken