expected accuracy on training audio set?

[goose]

Good morning,

warning - complete noob here.

I just started diving into sphinx a few days ago and managed to build my first
acoustic models today. To have some orientation where I am I would like to
know what accuracy i should expect if I apply the new model on parts of the
training set itself (Transcriber demo - Right now it is still quite poor but
there are a bunch of errors during the training ... )

Thanks

martin

[Nickolay V. Shmyrev]

Accuracy must be high. If you have issues, you need to fix the mistakes you
have made in the process.

Please also follow the tutorial precisely. You must get the accuracy estimate
with the decoder during training, not with the transcriber demo.

[goose]

thanks a lot nshmyrev, I sort of felt already that no point to move on while
accuracy on the training audio is poor - will dive into the tutorial again as
suggested and look at the docoder info which i wasnt aware of yet - really
getting fascinated about speech recognition