Communication support system that can increase the likability of dialogue partners using ChatGPT

Abstract

In recent years, the number of users of social networking services has been increasing, and with it, opportunities for communication with others. While this is a good thing because it makes it easy to communicate with others, messages and posts that hurt or offend others are on the increase. The offensive contents include not only objectively understandable contents such as slander and libel, but also contents that the other party may find offensive depending on the relationship between the two parties and the situation. There are few existing communication support tools that directly correct input comments. In this study, in addition to objective criteria, we use ChatGPT to extract features of sentences that the other party finds favorable or unfavorable from the dialogue history between the user and the other party. We aim to realize better communication by constructing a system that can propose sentence modification based on these features to the other party, if necessary.