I downloaded the speaker dependant training demo HTS-demo_Nitech-ATR503-m001 and
runned it under Red hat enterprise linux 5.0. First i configured the demo and
synthesized speech in japanese. After that,i replaced the data files in the data
directory such as fulllab,monolab,gen and raw with the files designed for other
language.At the start,there was errors like "too few observation squeences",i
found that it was the error on the mono lab files,i corrected that,and there was
new errors like "Error } expected ERROR [+7230] EdError: item list parse error".
The system generated file named m2f.hed. There was a misplaced new line in the
m2f.hed file. I am searching the Training.pl for the code that may caused the
errors. I found that the demo is speaker dependant,is that the different speaker
caused the errors?i guess it is the data files i replaced that caused the
error,but i can't find the problem,what is that?please help,thanks. the part of
the log file as follows: