Zero-shot Multi-Domain Dialog State Tracking Using Descriptive Rules

TL;DR

This paper introduces a framework that integrates descriptive logical rules into neural networks for dialog state tracking, enabling zero-shot learning of unseen labels without altering the network architecture.

Contribution

It proposes a novel method to incorporate logical rules into neural networks via loss function modifications, allowing zero-shot handling of unseen labels in dialog state tracking.

Findings

Logical rules improve zero-shot label prediction.

No degradation in existing system performance.

Applicable to neural dialog state trackers.

Abstract

In this work, we present a framework for incorporating descriptive logical rules in state-of-the-art neural networks, enabling them to learn how to handle unseen labels without the introduction of any new training data. The rules are integrated into existing networks without modifying their architecture, through an additional term in the network's loss function that penalizes states of the network that do not obey the designed rules. As a case of study, the framework is applied to an existing neural-based Dialog State Tracker. Our experiments demonstrate that the inclusion of logical rules allows the prediction of unseen labels, without deteriorating the predictive capacity of the original system.