The Chain of Questions for Dialogue Continuance

Abstract

In recent years, with the advent of large language models, there has been a growing interest in dialogue systems used as spoken agents, and communication robots are now becoming available for practical use. Much research has been conducted on non-task oriented dialogues, where the goal is for the dialogue system to interact naturally with a person. However, at present, there has not been much success in terms of the system being able to maintain a natural dialogue. In this study, based on the idea that questions from the system side are important, we focus on the chain of generated questions and propose a method that performs this chaining while taking into account the relevance of both the user's response and the dialogue. The experimental results show the effectiveness of our proposed method of the chain of generating questions using the related words.