Semantic systems improve information management by providing mechanisms to access objects by their attached semantic metadata elements rather than by their static location within a hierarchical structure. Semantic file systems apply these mechanisms to data stored in file systems in order to increase the quality of information retrieval. However, the effectiveness of semantic file systems depends strongly on the enrichment of files with relevant and comprehensive metadata, which is a tedious task if done manually. In this paper we present an approach that attains to disburden users from this expensive task by automatically generating useful metadata for files. It analyzes file system interactions and transforms them into contextual and semantic file relationship metadata, which can then be queried in order to search and retrieve information more efficiently. In this paper we give an overview on the algorithms we have used for this purpose, as well as an experimental evaluation of our prototype implementation, which demonstrates the effects of varying algorithm parameters.