Folksonomy systems enable users to participate in the Webcontent creation process by annotating (tagging) resourceswith freely chosen keywords. Still, it is an open issue howto exploit this user-created content, and how to processand use these emergent semantics effectively. We investigatehow the context of Web resources can be utilizedto improve recommender strategies in social tagging systems.We focus on the GroupMe! system, which enablesusers to create groups in order to bundle Web resources.GroupMe! groups form valuable context information forthe resources contained in such groups. In this paper we exploitgraph-based tag recommendation strategies, evaluatethem in the GroupMe! dataset, and benchmark our resultsagainst other approaches. In our evaluations we show thatgraph-based strategies outperform other approaches, andshow the immanent benefit of graph-based recommendationstrategies which exploit the group context for recommendingtags to untagged resources.