We're all Ontologists Now
Ian Bailey
Using the OWL language doesn’t magically turn a data structure into a model of what exists (i.e. an ontology). It requires a good deal of analysis to develop a proper ontology from a pre-existing data model, and it can’t really be done properly without access to existing data that conforms to that data model. There are techniques for this – we use the BORO Methodology, and other ontology development efforts have published development guidelines and FAQs. There are even data model improvement guides such as Shell’s Developing High Quality Data Models booklets from ten years ago that start to introduce some of the principles required when re-engineering a data model to an ontology. I didn’t see much of this in evidence in the body of work published on the web.
Taxonomies are developed on the principle of broader and narrower terms. However, a taxonomy generally doesn’t distinguish between classes and individuals. Nor do taxonomies distinguish between whole-part, type-instance and sub-supertype relationships. In a taxonomy, they’re all just broader-narrower relationships. As with a data model, one cannot simply turn a taxonomy into an ontology be representing it in OWL. OWL class hierarchies are based on sub-supertype relationships, which is just one case of broader-narrower, and one that only exists between classes at that.
del.icio.us