Miyata Lab / 宮田研究室

Department of Information Science, College of Humanities and Sciences, Nihon University / 日本大学 文理学部 情報科学科

Joker

Interactive agents are one of the technologies supporting the continuously evolving field of information technology. However, a certain number of users do not feel a sense of familiarity with impersonal conversations with these agents. To solve this problem, we propose an interactive agent that ‘deliberately misunderstands’ part of the user’s speech and asks for clarification. This approach is expected to introduce humor into the conversation, thereby increasing the user’s sense of familiarity. Our model is based on the incongruity-resolution model, which is supported in the field of cognitive science. In this model, the agent misunderstands words by outputting ones that sound similar to the original but have significantly different meanings.

Demo movie


Journal articles
  1. Kenro Go, Shunsuke Tokuda, Haruna Niiyama, Toshiki Onishi, Asahi Ogushi, and Akihiro Miyata: Evaluation of Prosodic Features Suitable for Conversational Agents Replying with a Joke. Journal of Information Processing, Vol.32, pp.35–40 (2024).
  2. Kenro Go, Toshiki Onishi, Asahi Ogushi, Shunsuke Tokuda, and Akihiro Miyata: Evaluation of Motivation for Conversational Agents Replying to Manzai-Style Jokes. Transactions of the Virtual Reality Society of Japan, Vol. 28. No. 1, pp.43–53 (2023).
International conferences
  1. Kenro Go, Toshiki Onishi, Asahi Ogushi, and Akihiro Miyata: Conversational Agents Replying with a Manzai-style Joke. Proc. 33rd Australian Conference on Human-computer-interaction (OzCHI ’21), pp.221–231 (2021).

Part of this project (-2017) was conducted with NTT Resonant Inc.

Back to the project list

Scroll to top