Does anyone know of software than can detect speech (speech without music present) in a large portion of audio, so as to save time finding it.  Are speech-related audio files sufficiently predictable for a program to accurately detect the presence of speech?  Thus, if I had 15 hours of audio of a rock festival on a particular stage or say 12 hours of a radio station, is there a way to locate all speech present so as to easily delete it?

I am no help here except bringing thread to the top. I believe that it exists, and at least govermental spies use sophisticated versions of it to manage tremendous amounts of information. Google is your helper.