Chat, Shift and Perform: Bridging the Gap between Task-oriented and Non-task-oriented Dialog Systems

TL;DR

CASPER is a novel dialog system that integrates chat, topic shifting, and task performance models to create more natural and satisfying conversations by seamlessly transitioning between open-domain and task-oriented dialogs.

Contribution

The paper introduces CASPER, a three-component dialog system that effectively bridges non-task and task-oriented conversations, demonstrating improved naturalness and user satisfaction.

Findings

CASPER outperforms baseline systems in naturalness and satisfaction.

Topic shifting via shifter enhances task-elicitation rates.

Removing shifter reduces dialog quality and task success.

Abstract

We propose CASPER (ChAt, Shift and PERform), a novel dialog system consisting of three types of dialog models: chatter, shifter, and performer. Shifter, which is designed for topic switching, enables a seamless flow of dialog from open-domain chat- to task-oriented dialog. In a user study, CASPER gave a better impression in terms of naturalness of response, lack of forced topic switching, and satisfaction compared with a baseline dialog system trained in an end-to-end manner. In an ablation study, we found that naturalness of response, dialog satisfaction, and task-elicitation rate improved compared with when shifter was removed from CASPER, indicating that topic shift with shifter supports the introduction of natural task-oriented dialog.