Last week, I ran sphinxTrain and adapted it to sphinx2. But coverting sphinx3 format models to sphinx2 format was not easy! Indeed:
* I'm not sure of the names of files (variables needed) I have many files (means files, matrices files, models files) and I don't know which one to choose to convert s3 format.
* makesendmp and mdef2phonemap don't exist. So I ran mk_s2sendump and mk_s2phonemap. Is it correct?
And when I wanted to test my new LM by running sphinx2 recognition engine, I got that error:
sc_vq.c(1584): **ERROR** Too low senone PDF value: -181248
Any idea???
Thank you very much.
Julie
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
Last week, I ran sphinxTrain and adapted it to sphinx2. But coverting sphinx3 format models to sphinx2 format was not easy! Indeed:
* I'm not sure of the names of files (variables needed) I have many files (means files, matrices files, models files) and I don't know which one to choose to convert s3 format.
* makesendmp and mdef2phonemap don't exist. So I ran mk_s2sendump and mk_s2phonemap. Is it correct?
And when I wanted to test my new LM by running sphinx2 recognition engine, I got that error:
sc_vq.c(1584): **ERROR** Too low senone PDF value: -181248
Any idea???
Thank you very much.
Julie
Hi Julie.
I can't undertand why do you find it so dificult. Have you tried to run make_s2_models.pl?
This script generates sphinx2 models automaticly!