SF-DST: Few-Shot Self-Feeding Reading Comprehension Dialogue State Tracking with Auxiliary Task

TL;DR

This paper presents SF-DST, a few-shot dialogue state tracking model that uses self-feeding belief states and an auxiliary slot-gate task to improve accuracy in multi-turn dialogues, achieving state-of-the-art results.

Contribution

Introduces an ontology-free few-shot DST model with self-feeding belief states and a novel slot-gate auxiliary task for better slot classification.

Findings

Achieved top scores in four domains on multiWOZ 2.0 in few-shot settings.

Enhanced dialogue understanding with self-feeding belief states.

Improved slot classification accuracy with the auxiliary task.

Abstract

Few-shot dialogue state tracking (DST) model tracks user requests in dialogue with reliable accuracy even with a small amount of data. In this paper, we introduce an ontology-free few-shot DST with self-feeding belief state input. The self-feeding belief state input increases the accuracy in multi-turn dialogue by summarizing previous dialogue. Also, we newly developed a slot-gate auxiliary task. This new auxiliary task helps classify whether a slot is mentioned in the dialogue. Our model achieved the best score in a few-shot setting for four domains on multiWOZ 2.0.