GOOG-411 created to train speech-to-text software

running the free service to build up a corpus of voices? neat idea, but seems inefficient