Support and Evaluation of Communication Skills in Group Discussions by Conversational Agents

2016 Fiscal Year Final Research Report

• Project/Area Number: 25280076
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Partial Multi-year Fund
• Section: 一般
• Research Field: Human interface and interaction
• Research Institution: Seikei University
• Principal Investigator: Yukiko Nakano 成蹊大学, 理工学部, 教授 (40422505)
• Co-Investigator(Kenkyū-buntansha): 岡田 将吾 東京工業大学, 総合理工学研究科(研究院), 助教 (00512261) ; 黄 宏軒 立命館大学, 情報理工学部, 准教授 (00572950) ; 高瀬 裕 成蹊大学, 理工学部, 助教 (60744000) ; 林 佑樹 成蹊大学, 理工学部, 助教 (40633524)
• Project Period (FY): 2013-04-01 – 2017-03-31
• Keywords: マルチモーダルインタラクション / 非言語行動認識

Outline of Final Research Achievements

This research project collected verbal and nonverbal behaviors in group discussions, and established models for estimating communication skills and important statements in the discussion. First, we conducted a corpus collection experiment using 40 subjects, and collected 14000 utterances, from which estimation models were created by exploiting machine learning techniques. The accuracy of the communication skill (high or low) estimation model was 93%. The model performance for detecting important utterances was 0.7 in F-measure. We also developed a conversation intervention robot and demonstrated its effectiveness.

Free Research Field

知的インタフェース