The final goal of this study is to create an online conversation support system that helps dialogues between two individuals aiming for a targeted dialogue in a goal-oriented conversation. This study designed a function to detect whether an online conversation between two individuals is suitable for the targeted dialogue. We assumed that speakers aimed for assertive communication and defined it as a targeted dialogue. We evaluated the suitability of conversations for the targeted dialogue using our original index, the targeted dialogue score. If the index value was lower than the threshold, we defined the dialogue as “Not assertive communication.” We developed a conversational activeness index to assess the suitability of conversations for targeted dialogues. We extracted the index from the voice data of conversations between individuals. We found a high correlation between the conversational activeness index score and the targeted dialogue score. We employed logistic regression analysis to determine the suitability of online conversations for targeted dialogues based on the correlation between the conversational activeness index score and the targeted dialogue score. The cross-validation of this model was 0.93, indicating high accuracy. However, false positives also occurred because the model was created only based on the conversational activeness index.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Explore the Timing of Conversational Interventions to Design a System that Supports Online One-On-One Discussions

  • Konosuke Ikeda,
  • Akihiro Ogino

摘要

The final goal of this study is to create an online conversation support system that helps dialogues between two individuals aiming for a targeted dialogue in a goal-oriented conversation. This study designed a function to detect whether an online conversation between two individuals is suitable for the targeted dialogue. We assumed that speakers aimed for assertive communication and defined it as a targeted dialogue. We evaluated the suitability of conversations for the targeted dialogue using our original index, the targeted dialogue score. If the index value was lower than the threshold, we defined the dialogue as “Not assertive communication.” We developed a conversational activeness index to assess the suitability of conversations for targeted dialogues. We extracted the index from the voice data of conversations between individuals. We found a high correlation between the conversational activeness index score and the targeted dialogue score. We employed logistic regression analysis to determine the suitability of online conversations for targeted dialogues based on the correlation between the conversational activeness index score and the targeted dialogue score. The cross-validation of this model was 0.93, indicating high accuracy. However, false positives also occurred because the model was created only based on the conversational activeness index.