Our aim is to create a corpus-based semantic and grammatical description of prepositional constructions using empiric corpus data. The methodology for processing corpus data and calculating frequency characteristics of prepositional constructions in modern Russian texts is presented. The prepositional values are described in terms of semantic rubrics based on the notion of syntaxemes introduced by G.Zolotova. We investigate prepositional constructions within the specific semantic rubrics: mediative and transitive.