Multi-Agent Task-Oriented Dialog Policy Learning with Role-Aware Reward Decomposition

ACL 2020