Learning through Dialogue Interactions by Asking Questions

TL;DR

This paper explores how dialogue agents can learn more effectively by asking questions, using a simulator and real-world experiments to demonstrate improved learning in interactive settings.

Contribution

It introduces a novel framework for training dialogue agents that learn from asking and answering questions, validated through synthetic tasks and human experiments.

Findings

Learners improve their performance when asking questions.

The proposed method benefits both offline and online reinforcement learning.

Real-world validation confirms the effectiveness of question-asking in dialogue agents.

Abstract

A good dialogue agent should have the ability to interact with users by both responding to questions and by asking questions, and importantly to learn from both types of interaction. In this work, we explore this direction by designing a simulator and a set of synthetic tasks in the movie domain that allow such interactions between a learner and a teacher. We investigate how a learner can benefit from asking questions in both offline and online reinforcement learning settings, and demonstrate that the learner improves when asking questions. Finally, real experiments with Mechanical Turk validate the approach. Our work represents a first step in developing such end-to-end learned interactive dialogue agents.