Towards Adaptive Spoken Dialog Systems