Leveraging Slot Descriptions for Zero-Shot Cross-Domain Dialogue State Tracking
arXiv:2105.04222
Abstract
Zero-shot cross-domain dialogue state tracking (DST) enables us to handle task-oriented dialogue in unseen domains without the expense of collecting in-domain data. In this paper, we propose a slot description enhanced generative approach for zero-shot cross-domain DST. Specifically, our model first encodes dialogue context and slots with a pre-trained self-attentive encoder, and generates slot values in an auto-regressive manner. In addition, we incorporate Slot Type Informed Descriptions that capture the shared information across slots to facilitate cross-domain knowledge transfer. Experimental results on the MultiWOZ dataset show that our proposed method significantly improves existing state-of-the-art results in the zero-shot cross-domain setting.
arXiv commentNAACL 2021
References in corpus (12)
- Decoupled Weight Decay Regularization
- A Simple Language Model for Task-Oriented Dialogue
- Multi-domain Dialogue State Tracking as Dynamic Knowledge Graph Enhanced Question Answering
- Transferable Multi-Domain State Generator for Task-Oriented Dialogue Systems
- CoCo: Controllable Counterfactuals for Evaluating Dialogue State Trackers
- Language Models as Few-Shot Learner for Task-Oriented Dialogue Systems
- MultiWOZ 2.2 : A Dialogue Dataset with Additional Annotation Corrections and State Tracking Baselines
- SUMBT: Slot-Utterance Matching for Universal and Scalable Belief Tracking
- Dialog State Tracking: A Neural Reading Comprehension Approach
- Efficient Dialogue State Tracking by Selectively Overwriting Memory
- From Machine Reading Comprehension to Dialogue State Tracking: Bridging the Gap
- Slot Self-Attentive Dialogue State Tracking