We will present a new minimally supervised machine learning framework for extracting relations of various complexity. Bootstrapping starts from a small set of n-ary relation instances as "seeds" in order to automatically learn pattern rules from parsed data, which then can extract new instances of the relation and its projections. We propose a novel rule representation model that enables the composition of n-ary relation rules on top of the rules for projections of the relation. The compositional approach to rule construction is supported by a bottom-up pattern extraction method working on dependency structures. In comparison to other automatic approaches, our rules cannot only localize relation arguments but also assign their exact target argument roles. The evaluation results compare favorably with those of existing pattern acquisition approaches in both recall and precision. For one extraction task a single seed event suffices to get patterns that find most of the relevant events. For another task we need larger number of seed events
in order to get a satisfactory performance. We use known results from graph theory to describe the relevant differences between extraction
domains and propose some strategies for improving the selection and acquisition of effective seeds.