The faceted search paradigm allows to intuitively categorize the search results along orthogonal facets. Such facets, for example, can group the results along authors, publication year, or topics in case of digital libraries. One open issue, however, is how facets themselves can be organized and presented, especially if a facet is highly dynamic and is too large to be presented at once. In this paper we demonstrate the Semantic GrowBag approach to automatically organize facets, namely a topic facet, for community-specific document collections. The approach uses the documents’ metadata tags to extract information about prevalent topics and then applies PageRank to determine intrinsic relations between these topics as supported by a collection’s documents. We discuss a short use case for topics extracted for computer science documents in the DBLP collection and show that we can even determine a topic’s development over time to enable a customizable organization of the topic facet. Given that facets have to be individually adapted to different users, user groups, or communities our algorithm is an important step towards automatic facet organization.