We present a novel problem that allows to infer a small and interpretable set of prevalent opinion ranges in signed graphs, that explain the users' interactions.