Topic Selection Using Conceptual Distance: How to Select Topics that are Interesting but Unfamiliar to Users

抄録

In this study, we established a topic selection method that recommends topics that are interesting and unfamiliar to users. To achieve this aim, we used conceptual distance to identify topics that were unfamiliar to users and improved the accuracy of this method by removing conceptually similar words. Many words used in conversations are excluded in the dictionaries and thesauruses. Thus, we developed a model for conceptual distance measurement using machine learning to measure conceptual distances even for such words. By conducting the subject experiments, we confirmed that the established system recommends topics a user is interested in but unfamiliar with compared with the baseline method developed in a previous research.