A method for conversational topic recommendation to appropriate user group

As intimacy in human relationships is getting reduced, social problems increase, and a number of people feel cut off. To support making friends and activating communication, we propose a topic recommendation method for conversation to an appropriate user group. Although previous methods have an issue where many of persons may not enjoy conversation by the topic recommendation method, our proposed method can provide the topics which are preferred by more persons for their conversation, and more persons can enjoy their conversation. The method utilizes the similarities in all combination between topic and a user preference. The topic preference vector which is formed by user as the element and the similarity as the value is constructed for each topic, and the topics are clustered by using the topic preference vectors, based on the similarities. Then, the user groups are extracted from the topic clusters, and the corresponding topics are recommended to the user groups. For the proposed method, the effectiveness of user group construction, the completeness of the user preference extraction and the appropriateness of topic selection are investigated by conducting an simple experiment. As the result, we confirmed the user group construction can help to have more enjoyable conversation, and although the user preference extraction was not enough, the topics offered by the proposed method and along with a category are appropriate for the known persons and the unknown persons, respectively.