i build acoustic model for Arabic language using my own dictionary ( i have
developed automatic pronunciation dictionary tool-kit)
training hours 5 hours
number of word 2274
number of speakers 6 all of them speaks the same 571 sentences
after training my model i tested it using
perl scripts_pl/decode/slave.pl ,
with different speaker who says the same 671 sentences using in training and
the results was as follows:
TOTAL Words: 2274 Correct: 1646 Errors: 723
TOTAL Percent correct = 72.38% Error = 31.79% Accuracy = 68.21%
is my training perfect?
how many traing hours do I need to get accuracy 99.9%
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
thank you Nickolay
i believe the problem is in my dictionary !!!!!!!!
for such vocabulary and for the same training hours what is expected error, to
say that my training is ok?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
i build acoustic model for Arabic language using my own dictionary ( i have
developed automatic pronunciation dictionary tool-kit)
training hours 5 hours
number of word 2274
number of speakers 6 all of them speaks the same 571 sentences
after training my model i tested it using
perl scripts_pl/decode/slave.pl ,
with different speaker who says the same 671 sentences using in training and
the results was as follows:
TOTAL Words: 2274 Correct: 1646 Errors: 723
TOTAL Percent correct = 72.38% Error = 31.79% Accuracy = 68.21%
is my training perfect?
how many traing hours do I need to get accuracy 99.9%
I suppose not
It's impossible to get such high accuracy. Even very simple and long-studied
configs like digits recognition are running at 99.6%.
thank you Nickolay
i believe the problem is in my dictionary !!!!!!!!
for such vocabulary and for the same training hours what is expected error, to
say that my training is ok?