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
This paper proposes a method to construct a character dialogue system using pseudo-persona dialogue data created from novels, and reports the evaluation results of the constructed system.In order to create dialogue system that reflects the character’s personality, we first manually extracted the persona information and the character’s utterance from novel. Since the extracted character’s utterances are not dialogue form and cannot be used for training a dialogue model, we created pseudo-dialogue response pairs by creating a pre-utterance generation model.By training from personas and pseudo-response pairs, we constructed a character dialogue system. The results of user evaluations of the constructed models showed the models of the proposed method received highly recommended, suggesting that the proposed method is effective.