The usefulness of a recommendation is determined by its accuracy. However, users are not always satisfied with only accuracy. The usefulness of a recommendation system is determined not only by its ability to recommend an item that the user knows but also by its ability to recommend one that the user does not know but may like. This means that the user's taste tendencies should be taken into account. We have developed an algorithm that extracts a topic from a user's rating history automatically and then diversifies the topics on the list of recommendations presented to the user. Experiments showed that the level of user satisfaction is higher with such a list than one created using collaborative filtering.
抄録全体を表示