If you say taxa are individuals and I say they are classes we still need a term that is the thing we dissagree about.

I believe people will learn to regret generating large semantic networkS with SKOS or OBO when they realise that they can not be used for inference because they are too vague.

If, as I worry, your transformations are going from the more to less specific, I think this could account for the concern Donald raises that modeling properties becomes particularly difficult, because one’s intuition gets stuck in the specifics. As I come to understand more about the uses scientific communities are making of Knowledge Representation tools, I am struck by how frequently a community neglects to–or at least has difficulty to– separate thesauri from knowledge models. Since the main predicate in most thesauri is essentially “moreSpecificThan” (e.g. skos:narrowerThan), mainly what you get out of thesauri is the ability to do not-stupid data integration. Important as that is, I’m not content with building systems for that, because most such systems can be built with good code generation tools such as databinding frameworks. This year, the large pine tree in my yard (a.k.a “my garden”) dropped perhaps 20 times more cones than in any year in my 30 year residence here. I’m only a little interested in knowing how geographically wide spread this is; I’m only a little interested in learning that there is a term for this phenomenon (“mast year”); I’m a lot interested in learning \why/ this happens, and I want to build information systems that allow amateurs and scientists to offer hypotheses to an information system which then goes out to discover rules, relations, and data that will be determined to be either consistent or inconsistent with those hypotheses. I’m pretty sure that this requires correctly representing what is general and what is specific, and not just at the term level.

With regard to the flexibility, however, I believe the better management system might be semantic mediawiki. SMW allows humans to define the terms in a satisfying way (rich multimedia text), and ontology experts can map and relate these with all necessary classes and properties, in a very intuitive way to biologists. All information is automatically exposed as OWL/RDF – and could be harvested through a simple job.

I imagine we would tend to have a vocabulary per class. TaxonName has its own namespace and so would contain both the class term and a term for each of the properties.

Thanks for this – it’s a very nice move and I support our exploring this approach. My only real question is how much of the TDWG Ontology work can really be done this way. It is an excellent way for us to build a list of entities – with the ontology classes being one such list. It does not seem to give us a way to define the set of properties associated with those classes, which is the largest part of our modeling task – or am I missing something?

Donald