[4Rin1] Interactive 2

[4Rin1-91] Multivariate Analysis between User Interests and Article Summarization Strategies for Conversational News Contents Delivery

〇Mayu Okuda¹, Hiroaki Takatsu¹, Yoichi Matsuyama¹, Hiroshi Honda², Shinya Fujie^1,3, Tetsunori Kobayashi¹ (1.Waseda University, 2.Honda R&D Co.,Ltd. , 3.Chiba Institute of Technology)

Keywords:User Model, Automatic Text Summarization, Spoken Dialogue System

In this paper, we analyze the relationship between the user's interest and the article summarization strategy for conversational news article transmission. We have proposed a system that provides information necessary for users through conversational interaction and prevents unnecessary information in order to efficiently transmit information such as news articles. In a conversation system that conveys such a large amount of information, it is important to adjust the amount of information to be conveyed according to the user's interest in efficient information transmission. Therefore, we conducted two experiments and performed a multivariate analysis to investigate the relationship between user interests and article summarization strategies. The subjects were Waseda University students and they were asked about the profile, such as interest of genres. In addition, we showed them three summaries of different lengths automatically generated by the system and let them answer their favorite rankings. Furthermore, we asked them to answer their needs for each sentence in the news articles. Using these datasets, we showed that there is a relationship between the user's interest and the amount of summaries they prefer, and showed the effect of changing the amount of summaries provided for each user and for each article.