Recommender systems rely on the opinions of a community of users to provide "recommendations" that can help users of the same community in discerning content of interest from a wide range of possibilities. Particularly, collaborative information filtering represents one of techniques widely exploited by recommender systems to suggest which items better meet the user needs and preferences. This paper introduces a model for collaborative filtering based on Formal Concept Analysis, a theoretical framework suitable to generate correlations among data through a lattice design. In particular, a fuzzy annotation of the lattice allows discovering similarities among items as well as users, arranged as a ranked list.