Currently, news contents are based on traditional taxonomy with hierarchical structure. Meanwhile, social bookmarks have a flat structure called folksonomy which has less difficulty to get news rather than taxonomy. However, bookmark tags depends on users. It causes three problems; polysemy, synonymy, difference of understanding. In this study, we propose a new method to recommend interesting news by using social bookmarking. It uses tags attached by a group of users who have similar preferences. Although this method still uses tagging system, the tags represent same understanding of the group. It reduces the influence of the three problems above. We also employ thesaurus to extrapolate sparse descriptions. We had evaluated appropriateness assigned to the web pages by the users of articles selected and recommended for the web pages. In our experiment, social bookmarking service, called 4dk, is used. 4dk is based on the group. All tags attached in 4dk are assumed as tags attached by users who have same preferences. We showed high precision of the set of recommended pages through statistical tests such as wilcoxon test and sigmoid-like function test developed in our study.