Topic Recommendation Method related to a Present Topic for Continuing a Conversation

Abstract

This paper proposes a topic recommendation method. The methods recommends next topics related to a present topic. The purpose of the proposed method is to support to continue conversations. The proposed method detects spoken words of two users, makes a log text of the spoken words. Then, the method extracts a present topic from the log text, and evaluates relations between the present topic and next topic candidates. The method chooses next topics and recommends them to the users. The proposed method has been realized as a conversation support system that has an interface of Head Mounted Display (HMD) showing the log text, a word of the present topic, and words of the next topic. One of the users of the system wearing the HMD can grasp what topic is talked, and can choose a next topic from recommended them. The users can continue a conversation by talking about recommended topics. Experimental results showed that the realized system could support to continue conversations. We also found that the realized system supported users to obtain equal chances to speak.