MultiWOZ 2.4: A Multi-Domain Task-Oriented Dialogue Dataset with Essential Annotation Corrections to Improve State Tracking Evaluation

The MultiWOZ 2.0 dataset was released in 2018. It consists of more than 10,000 task-oriented dialogues spanning 7 domains, and has greatly stimulated the research of task-oriented dialogue systems. However, there is substantial noise in the state annotations, which hinders a proper evaluation of dialogue state tracking models. To tackle this issue, massive efforts have been devoted to correcting the annotations, resulting in 3 improved versions of this dataset (i.e., MultiWOZ 2.1-2.3). Even so, there are still lots of incorrect and inconsistent annotations. This work introduces MultiWOZ 2.4, in which we refine all annotations in the validation set and test set on top of MultiWOZ 2.1. The annotations in the training set remain unchanged to encourage robust and noise-resilient model training. We further benchmark 8 state-of-the-art dialogue state tracking models. All these models achieve much higher performance on MultiWOZ 2.4 than on MultiWOZ 2.1.

Slot Self-Attentive Dialogue State Tracking

An indispensable component in task-oriented dialogue systems is the dialogue state tracker, which keeps track of users' intentions in the course of conversation. The typical approach towards this goal is to fill in multiple pre-defined slots that are essential to complete the task. Although various dialogue state tracking methods have been proposed in recent years, most of them predict the value of each slot separately and fail to consider the correlations among slots. In this paper, we propose a slot self-attention mechanism that can learn the slot correlations automatically. Specifically, a slot-token attention is first utilized to obtain slot-specific features from the dialogue context. Then a stacked slot self-attention is applied on these features to learn the correlations among slots. We conduct comprehensive experiments on two multi-domain task-oriented dialogue datasets, including MultiWOZ 2.0 and MultiWOZ 2.1. The experimental results demonstrate that our approach achieves state-of-the-art performance on both datasets, verifying the necessity and effectiveness of taking slot correlations into consideration.