Ontological Analysis of Taxonomic Relationships

Abstract

Taxonomies based on a partial-ordering relation commonly known as is-a, class inclusion or subsumption have become an important tool in conceptual modeling. A well-formed taxonomy has significant implications for understanding, reuse, and integration, however the intuitive simplicity of taxonomic relations has led to widespread misuse, making clear the need for rigorous analysis techniques. Where previous work has focused largely on the semantics of the is-a relation itself, we concentrate here on the ontological nature of the arguments of this relation, in order to be able to tell whether a single is-a link is ontologically well-founded. For this purpose, we discuss techniques based on the philosophical notions of identity, unity, essence, and dependence, which have been adapted to the needs of information systems design. We demonstrate the effectiveness of these techniques by taking real examples of poorly structured taxonomies, and revealing cases of invalid generalization. The result of the analysis is a cleaner taxonomy that clarifies the modeler’s ontological commitments.