Action State Update Approach to Dialogue Management

TL;DR

This paper introduces the Action State Update (ASU) approach for dialogue management, which uses a binary classifier to interpret user utterances and update dialogue states without relying on domain-specific natural language understanding components.

Contribution

The paper presents a novel ASU method employing a statistically trained classifier and active learning to interpret referring expressions in dialogue systems.

Findings

ASU effectively interprets user utterances including referring expressions

Active learning improves training efficiency for dialogue state update

ASU performs well in both simulated and human evaluations

Abstract

Utterance interpretation is one of the main functions of a dialogue manager, which is the key component of a dialogue system. We propose the action state update approach (ASU) for utterance interpretation, featuring a statistically trained binary classifier used to detect dialogue state update actions in the text of a user utterance. Our goal is to interpret referring expressions in user input without a domain-specific natural language understanding component. For training the model, we use active learning to automatically select simulated training examples. With both user-simulated and interactive human evaluations, we show that the ASU approach successfully interprets user utterances in a dialogue system, including those with referring expressions.