Predictive Speech Recognition and End-of-Utterance Detection Towards Spoken Dialog Systems