The growth of user-generated content in quantity and quality has changed the way people use digital services, including geo-services. The process of map generalization is not an exception to this phenomenon. Earlier research has considered user-generated content as data sources for the generalization process. However, little work has been accomplished to date considering the knowledge that may be extracted from those sources, in particular from special place-related semantics captured in user-contributed feature tags. This study considers doing so from the perspective of folksonomies, presenting some first steps in that direction. In particular, this short paper shows, on the example of OpenStreetMap, how different similarity measures can be used to exploit folksonomy-based semantics in map generalization. And it shows how these semantics can be used to introduce behaviour changes in generalization operators, in particular in the selection and aggregation operators, respectively.