Budgeted Policy Learning for Task-Oriented Dialogue Systems
This addresses the challenge of efficient policy learning for dialogue systems with constrained user interactions, representing an incremental advancement.
The paper tackles the problem of learning task-oriented dialogue agents with a limited number of user interactions by extending Deep Dyna-Q with Budget-Conscious Scheduling, resulting in significant improvements in success rate over state-of-the-art baselines under a fixed budget.
This paper presents a new approach that extends Deep Dyna-Q (DDQ) by incorporating a Budget-Conscious Scheduling (BCS) to best utilize a fixed, small amount of user interactions (budget) for learning task-oriented dialogue agents. BCS consists of (1) a Poisson-based global scheduler to allocate budget over different stages of training; (2) a controller to decide at each training step whether the agent is trained using real or simulated experiences; (3) a user goal sampling module to generate the experiences that are most effective for policy learning. Experiments on a movie-ticket booking task with simulated and real users show that our approach leads to significant improvements in success rate over the state-of-the-art baselines given the fixed budget.