Reinforcement Learning for Adaptive Dialogue Systems