Stabilized In-Context Learning with Pre-trained Language Models for Few Shot Dialogue State Tracking
This work addresses the problem of unstable and inefficient in-context learning for dialogue state tracking, which is incremental as it builds on existing prompt-based methods with specific optimizations.
The paper tackled the instability and input length challenges of prompt-based methods for few-shot dialogue state tracking (DST) by adapting meta-learning, improving example retrieval, and using saliency to shorten dialogue text, achieving competitive results on MultiWOZ.
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.