UNO-DST: Leveraging Unlabelled Data in Zero-Shot Dialogue State Tracking
This work addresses the challenge of dialogue state tracking for AI systems in new domains without labeled data, representing an incremental advancement over previous zero-shot methods.
The paper tackled the problem of zero-shot dialogue state tracking by leveraging unlabelled data in target domains to transform it into few-shot scenarios, resulting in an 8% improvement in average joint goal accuracy across all domains in MultiWOZ.
Previous zero-shot dialogue state tracking (DST) methods only apply transfer learning, ignoring unlabelled data in the target domain. We transform zero-shot DST into few-shot DST by utilising such unlabelled data via joint and self-training methods. Our method incorporates auxiliary tasks that generate slot types as inverse prompts for main tasks, creating slot values during joint training. Cycle consistency between these two tasks enables the generation and selection of quality samples in unknown target domains for subsequent fine-tuning. This approach also facilitates automatic label creation, thereby optimizing the training and fine-tuning of DST models. We demonstrate this method's effectiveness on general language models in zero-shot scenarios, improving average joint goal accuracy by 8% across all domains in MultiWOZ.