Improving Recommendation Diversity by Maximizing Semantic Volume

The amount of information transmitted via electronic media is increasing, and recommender systems are being introduced. However, it has been pointed out that there are problems such as filter bubbles and echo chambers, which provide users with biased information due to excessive recommendations. In previous studies, we have discussed changes in user behavior based on the diversity of articles. In this study, we propose a recommender system that introduces a mechanism to improve the linguistic diversity of articles, and show through experiments that the system is able to recommend a variety of articles. We also have clarified the condition that merely displaying a variety of articles is not sufficient to improve the diversity of articles clicked by users, and that it is necessary to recommend articles that capture the interests of users.