On a Serendipity-oriented Recommender System based on Folksonomy and its Evaluation

The present paper proposes a recommendation method that focuses not only on predictive accuracy but also serendipity. In many of the conventional recommendation methods, items are categorized according to their attributes (genre, author, etc.) by the recommender in advance, and recommendations are made using the categorization. In the present study, the impression of users regarding an item is adopted as its feature, and items are categorized according to this feature. Such impressions are derived using folksonomy. A recommender system based on the proposed method was developed in the Java language, and the effectiveness of the proposed method was verified through recommender experiments.