Grounding Description-Driven Dialogue State Trackers with Knowledge-Seeking Turns

TL;DR

This paper introduces a method to improve dialogue state tracking models by grounding them in knowledge-seeking turns from dialogue data, enhancing robustness and accuracy without costly paraphrase augmentation.

Contribution

The authors propose incorporating knowledge-seeking turns into prompts during training and inference to improve schema robustness in dialogue state trackers.

Findings

Significant accuracy improvements on SGD and SGD-X datasets

Enhanced robustness to schema variations

Reduced need for costly paraphrase data augmentation

Abstract

Schema-guided dialogue state trackers can generalise to new domains without further training, yet they are sensitive to the writing style of the schemata. Augmenting the training set with human or synthetic schema paraphrases improves the model robustness to these variations but can be either costly or difficult to control. We propose to circumvent these issues by grounding the state tracking model in knowledge-seeking turns collected from the dialogue corpus as well as the schema. Including these turns in prompts during finetuning and inference leads to marked improvements in model robustness, as demonstrated by large average joint goal accuracy and schema sensitivity improvements on SGD and SGD-X.