Curiosity-Driven Reinforcement Learning For Dialogue Management

In this paper we describe the use of curiosity rewards for dialogue policy learning of goal oriented dialogues via reinforcement learning. Using curiosity improves state-action space exploration and helps overcome reward sparsity. Additionally, for goal oriented dialogues it makes sense to perform inherently curious actions in order to gain knowledge about the user goal. We show that intrinsic curiosity rewards can replace random epsilon-greedy exploration and stabilize training. The best results are achieved when curiosity rewards are combined with epsilon-greedy exploration.