Hi yossarian,

>50 hours or so of recorded audio from university lectures in wma format.

The easiest way to create acoustic models is with transcribed speech (i.e. the text is included with the speech) so that the training process can match the textual words to the speech sounds. Transcribing speech is *very* time consuming, and for this reason, we don't usually accept untranscribed speech. 

Is the speech contained in these lectures transcribed?  Is there any Copyright attached to the speech recordings?

thanks,

Ken

--- (Edited on 5/3/2009 8:28 pm [GMT-0400] by kmaclean) ---