A New Method for the Exploitation of Speech Recognition Systems