Speech technology in real world environment: early results from a long term study