Compositional reasoning aims to improve scalability of verification tools by reducing the original verification task into subproblems. The simplification is typically based on assume-guarantee reasoning principles, and requires user guidance to identify appropriate assumptions for components. In this paper, we propose a fully automated approach to compositional reasoning that consists of automated decomposition using a hypergraph partitioning algorithm for balanced clustering of variables, and discovering assumptions using the L* algorithm for active learning of regular languages. We present a symbolic implementation of the learning algorithm, and incorporate it in the model checker NUSMV. In some cases, our experiments demonstrate significant savings in the computational requirements of symbolic model checking.