Grounding Description-Driven Dialogue State Trackers with Knowledge-Seeking Turns
This work addresses robustness issues in dialogue state tracking for conversational AI systems, but it is incremental as it builds on existing schema-guided methods.
The paper tackled the problem of schema-guided dialogue state trackers being sensitive to schema writing style, and the result was improved model robustness with large average joint goal accuracy and schema sensitivity improvements on SGD and SGD-X datasets.
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.