A Study on Concept Acquisition Method for User Preference in Dialogue for Empathy

Abstract

The purpose of this research is to promote users' motivation to talk with robots. In order to improve the users' motivation in dialogue, it is necessary not only to generate empathic utterances of the robots but to show that the robots understand about target items. In this paper, we explore the method to acquire the concept of the users' preferences (likes and dislikes). In the proposed method, data on preference and similarity for items are prepared for each attribute of preference. In addition, we also organize the rules for estimating preferences and similarity with fewer observed data. As future work, it is necessary to verify whether or not the dialog robots with the proposed method can improve the users' satisfaction for the robots’ empathic utterances and the users' dialogue motivation.