In this paper, we propose a knowledge representation system that combines a greater expressive richness than that of the boolean or attributes frameworks, and an analogous treatment complexity. This system can be used in symbolic-numerical Clustering Analysis and will be used to build a Galois (concept) lattice.Data and data classes are described by labelled, recursive trees called genterms. The set of these are generated with an algebraic signature. In addition, it is ordered by a generalization relation inducing a semilattice and easily extending to a lattice. Genterm node labels belong to numerical or symbolic sets of type. Each of these sets is a semilattice in which the order relation can also be viewed as a generalization relation.The used formalism is similar to those of Abstract Data Type Signatures and offers an alternative to the Description Logic formalisms.The elements of the Galois lattice between a genterm lattice and a power set of objects are couples (concepts) containing one set of objects and its description by a unique genterm.