Social Networking Services (SNSs) have been gaining popularity in recent years. However, as the number of users of SNS increases, users are forced to spend a great deal of time in order to fi nd a good friend. The user recommendation system is a good tool that solves this problem, but in that case, the system should understand the user's interests. In the present paper, as a means of extracting a user interest, we construct a hierarchy of words from social bookmarking tags and emphasize characteristic words based on this relation. Then, we evaluate and demonstrate whether it is useful to extract an SNS user's interest.