Résumé
In many situations, one faces the need for integrating a set of conceptual hierarchies that represent parts of the same domain. The target structure of the integration is a unique conceptual hierarchy that embeds at least the total of the knowledge encoded in the initial ones. The key issues to address here are the discovery of higher-level abstractions on top of the existing concepts and the resolution of naming conflicts among these. Formal concepts analysis (FCA) is a mathematical approach toward abstracting from attribute-based object descriptions which has been recently extended to fit relational descriptions thus giving rise to the relational concept analysis (RCA) framework. Building up on RCA, our integration approach amounts to encoding the initial hierarchies into set of binary tables, one for each component category, e.g., classes, associations, etc., and subsequently constructing, querying and reorganizing the corresponding abstraction hierarchies until a unique satisfactory hierarchy is obtained. This paper puts the emphasis on the mechanisms of discovering new abstractions and exporting them between the abstraction hierarchies of related component categories. The impact of naming conflicts within the RCA process is discussed as well. The paper uses UML as description language for conceptual hierarchies.