sphinxtrain WARN: "corpus.c" messages

[Ruben Suarez]

Hello

I compiled myself voxforge_es_sphinx based on voxforge-es-0.2.tar.gz (availabe in download section) and latest wav's from voxforge.

When compiling I see lot of WARN messages in logidr/.bw.log like:

WARN: "corpus.c", line 1481: Utterance id in the transcription file, 'AGOs-20100831-jdm/wav/es-0015', does not match id in the control file, 'es-0015'.

But if I look at voxforgeessphinx.transcription.train there is:

<s> de unas parras artificiales cuyas hojas parecían retazos de terciopelo </s> (AGOs-20100831-jdm/wav/es-0015)

And in voxforgeessphinx.fields.train there is:

AGOs-20100831-jdm/wav/es-0015

It seems that id's match so, why the warning message? Is it on pourpouse? Is there any thing wrong with the id's?

Thanks!

[Nickolay V. Shmyrev]

This warning can be ignored.

Ideally the ID in transcription file must be the last part of control filename, not the whole part. So it should be:

  <s> de unas parras artificiales cuyas hojas parecían retazos de terciopelo </s>  (es-0015)

We probably want to update the preparation code to prepare proper transcription files.

[Ruben Suarez]

Ok, thank you for the info Nickolay V. Shmyrev

[Nickolay V. Shmyrev]

I have just fixed this warning in latest version, now it should proceed without a problem.

[Ruben Suarez]

One more question... When you say latest you mean latest on github repo or latest release?

I'm currently using 01-24-2016 release and the warnings are there. Also checked github repo and seems last commit is from that date too.

Thank you.

[Nickolay V. Shmyrev]

Our main repository is subversion here on sourceforge.

[Ruben Suarez]

Thanks! :)