Model-based Bayesian Reinforcement Learning for Dialogue Management

Pierre Lison

Reinforcement learning methods are increasingly used to optimise dialogue policies from experience. Most current techniques are model-free: they directly estimate the utility of va…