Development of Character-like Chatbot with Role Play-Based Question Answering and a Large Language Model

Abstract

We developed a chatbot that imitates a specific person by combining a large amount of dialogue data of that person with a large language model. Furthermore, we investigated the current performance and problems of the chatbot by conducting an open experiment and error analysis of the chatbot. The experiment confirmed the high naturalness and characterness of the chatbot. In addition, our investigation revealed that the errors specific to a person can be divided into two types: errors in attributes † 日本電信電話株式会社, NTT Corporation † † 現所属：rinna 株式会社, Present affiliation: rinna Co., Ltd. † † † NTT でのインターンの成果．現在，筑波大学修士課程修了, Work carried during internship at NTT. Currently, he completed the master's program at Tsukuba University. † † † † 株式会社ドワンゴ, DWANGO Co., Ltd. † † † † † 現所属：三菱ケミカルグループ株式会社, Present affiliation: Mitsubishi Chemical Group Corporation 本研究は IWSDS2022 の予稿集における "Fine-tuning a pre-trained Transformer-based encoder-decoder model with user-generated question-answer pairs to realize character-like chatbots"，および，ACL2022 の予稿集 に おける "Investigating person-specific errors in chat-oriented dialogue systems" をもとに翻訳したものである．