Host: The Japanese Society for Artificial Intelligence
Name : The 36th Annual Conference of the Japanese Society for Artificial Intelligence
Number : 36
Location : [in Japanese]
Date : June 14, 2022 - June 17, 2022
In this study, we propose a method of Tsukomi which is one of Japanese humor, to recover from the state of dialogue breakdown to the original dialogue flow more naturally in Chat-oriented dialogue system. Specifically, in addition to the chat dialogue system and the user, we prepare a third system that acts as a Tsukomi. The third system corrects the broken dialogue flow and recovers the broken dialogue by pointing out the mistakes in the dialogue. In order to verify the effectiveness of the proposed method, we conducted a dialogue evaluation experiment by incorporating the proposed method into an existing chat dialogue system and having users actually interact with the system. The results of the evaluation experiments showed that the proposed method was highly evaluated compared to the existing system without the proposed method and the system with the existing method. We also discuss the improvements of the proposed method using the dialogue logs obtained in the experiments.