Stabilized In-Context Learning with Pre-trained Language Models for Few Shot Dialogue State Tracking

TL;DR

This paper introduces a stabilized in-context learning approach for dialogue state tracking using pre-trained language models, employing meta-learning, improved retrieval, and a saliency model to enhance few-shot performance on MultiWOZ.

Contribution

It proposes a novel combination of meta-learning, improved retrieval, and dialogue length control to stabilize and enhance few-shot dialogue state tracking with PLMs.

Findings

Achieves highly competitive few-shot DST results on MultiWOZ.

Stabilizes model performance across different prompts.

Enables inclusion of more exemplars through dialogue length reduction.

Abstract

Prompt-based methods with large pre-trained language models (PLMs) have shown impressive unaided performance across many NLP tasks. These models improve even further with the addition of a few labeled in-context exemplars to guide output generation. However, for more complex tasks such as dialogue state tracking (DST), designing prompts that reliably convey the desired intent is nontrivial, leading to unstable results. Furthermore, building in-context exemplars for dialogue tasks is difficult because conversational contexts are long while model input lengths are relatively short. To overcome these issues we first adapt a meta-learning scheme to the dialogue domain which stabilizes the ability of the model to perform well under various prompts. We additionally design a novel training method to improve upon vanilla retrieval mechanisms to find ideal in-context examples. Finally, we introduce a saliency model to limit dialogue text length, allowing us to include more exemplars per query. In effect, we are able to achieve highly competitive results for few-shot DST on MultiWOZ.