Reward-Balancing for Statistical Spoken Dialogue Systems using Multi-objective Reinforcement Learning
This work addresses reward balancing for statistical spoken dialogue systems, which is an incremental improvement in dialogue policy optimization.
The paper tackles the problem of balancing multiple reward components in statistical spoken dialogue systems by proposing a structured method to find optimal reward component weightings using multi-objective reinforcement learning, which reduces the required training dialogues and is applied to six domains with comparisons to a baseline.
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.