2023 Fiscal Year Final Research Report

Development of a chat dialog system with personality and its impact on communication

Research Project

PDF

Project/Area Number	19K12191
Research Category	Grant-in-Aid for Scientific Research (C)
Allocation Type	Multi-year Fund
Section	一般
Review Section	Basic Section 61060:Kansei informatics-related
Research Institution	Tokyo University of Technology
Principal Investigator	Iwashita Shino 東京工科大学, コンピュータサイエンス学部, 准教授 (00360503)
Co-Investigator(Kenkyū-buntansha)	杉本徹芝浦工業大学, 工学部, 教授 (30277280) 伊藤紀子同志社大学, 文化情報学部, 准教授 (00391863) 林篤司国立研究開発法人農業・食品産業技術総合研究機構, 農業ロボティクス研究センター, 特別研究員 (60777895)
Project Period (FY)	2019-04-01 – 2024-03-31
Keywords	雑談対話システム / 個性 / 性格特性
Outline of Final Research Achievements	The purpose of this study was to construct a chat dialog system with personality. We conducted an experiment in which a person and an existing chatbot engaged in a chat dialogue (chat), and created and published a corpus of dialogues for 72 people to which the personality tags of the speaker and the JAIST tags related to the dialogue act were assigned. As a result of analyzing the relationship between the speaker's personality and the characteristics of his/her speech, we confirmed that the personality of the speaker as seen by others, such as the user, is directly related to the impression of the system, rather than the speaker's personality as perceived by the speaker himself/herself. By fine-tuning a pre-trained language model using the dialogue corpus as training data, we constructed a model for generating responses to chat dialogues, and were able to reflect the characteristics of personality impressions in terms of social skills and shyness.
Free Research Field	自然言語処理，感性情報処理
Academic Significance and Societal Importance of the Research Achievements	本研究では，個性を持つ雑談対話システムを構築することと，そのために必要な個性の特徴を紐づけした対話コーパスを構築することを目的とした．話者の性格特性と対話行為に関するJAISTタグを付与した72名分の対話コーパスを作成して公開することで，対話システムにおける性格の付与を目指す研究に貢献した．また，チャットベースの対話において，自覚している性格と表出された性格が異なり，表出された性格の方が重要であることを示したことで，対話システムに個性を付与する際の着目点を示すことができた．さらに，性格特性の一部を対話システムに付与することができ，ユーザに対して相性の良い雑談対話システムの実現に寄与できた．