The availability of semantically enriched data in repositories of larger content providers offers a means for new ways of multimedia presentation authoring. Existing (semi-)automatic content composition environments explore limited numbers of genres and concentrate on template-based approaches for composing discourse structures for these genres. We analyze a number of genres based on identifiable genre characteristics. We show that existing template-based approaches to content composition support the essay and biography genres that belong to one characteristic genre category but fail in supporting genres of another category. We present our approach to overcome this limitation and apply this approach to composing newspaper articles in the domain of fine arts