Continuous speech understanding based on automatic learning of acoustic and semantic models