Reinforcement learning is widely used for dialogue policy optimization where the reward function often consists of more than one component, e.g., the dialogue success and the dialogue length. In this work, we propose a structured method for finding a good balance between these components by searching for the optimal reward component weighting. To render this search feasible, we use multi-objective reinforcement learning to significantly reduce the number of training dialogues required. We apply our proposed method to find optimized component weights for six domains and compare them to a default baseline.
@article{arxiv.1707.06299,
title = {Reward-Balancing for Statistical Spoken Dialogue Systems using Multi-objective Reinforcement Learning},
author = {Stefan Ultes and Paweł Budzianowski and Iñigo Casanueva and Nikola Mrkšić and Lina Rojas-Barahona and Pei-Hao Su and Tsung-Hsien Wen and Milica Gašić and Steve Young},
journal= {arXiv preprint arXiv:1707.06299},
year = {2017}
}