Speech understanding systems: Final report