Improving Longer-range Dialogue State Tracking
This work addresses the problem of handling distracting contexts in longer dialogues for task-oriented dialogue systems, representing an incremental improvement in domain-specific DST.
The paper tackled the challenge of dialogue state tracking in longer conversations by proposing a model with hierarchical slot prediction, balanced training, and data perturbation, achieving improved performance on the MultiWOZ benchmark, especially for longer dialogues.
Dialogue state tracking (DST) is a pivotal component in task-oriented dialogue systems. While it is relatively easy for a DST model to capture belief states in short conversations, the task of DST becomes more challenging as the length of a dialogue increases due to the injection of more distracting contexts. In this paper, we aim to improve the overall performance of DST with a special focus on handling longer dialogues. We tackle this problem from three perspectives: 1) A model designed to enable hierarchical slot status prediction; 2) Balanced training procedure for generic and task-specific language understanding; 3) Data perturbation which enhances the model's ability in handling longer conversations. We conduct experiments on the MultiWOZ benchmark, and demonstrate the effectiveness of each component via a set of ablation tests, especially on longer conversations.