Formal Concept Analysis is an unsupervised clustering technique that accepts as inputs a set of objects with their corresponding attributes (called a formal context) and produces a mathematical lattice containing sets of objects and attributes at each position in the lattice. Each point in the lattice is called a formal concept representing the set of objects sharing the same values for a certain set of attributes; each sub-concept in the lattice is a subset of the objects of the concepts above it and a super-set of the attributes above it. This paper concerns generating meaningful natural language fragments from the list of attributes of a formal concept using a predicate map. It therefore address the important problem of determining natural language context in formal concept analysis.