Abstract: In this talk, I will introduce a new algorithm, Coherent Set Mining (CSM), for extracting sets of associated variables from binary data. This method relies on a new notion of object association, coherence, and an iterative testing approach based on asymptotic distribution results. In particular, CSM is designed to be effective in the common but challenging setting of non-identically distributed samples. I will share applications of CSM in text mining and in music recommendation systems.