- From: David Booth <david@dbooth.org>
- Date: Fri, 02 Oct 2009 14:38:50 -0400
- To: Toby Inkster <tai@g5n.co.uk>
- Cc: Kevin Tyson <kevin.tyson@gmail.com>, semantic-web@w3c.org, Bernard Vatant <bernard.vatant@mondeca.com>
True, but there's more to the story . . . 2009/10/2 Toby Inkster <tai@g5n.co.uk> > If I define: > > :Primate owl:equivalentClass [ > owl:unionOf (:Human :Monkey) > ] . > :Human owl:disjointFrom :Monkey . > > (And I realise I'm vastly oversimplifying the biology here!) > then a > statement like this: > > :Bob rdf:type :Primate . > > is not inconsistent. In the first snippet, you've asserted > that all > primates must be human or monkey, and cannot be both. In the > second, you've asserted that Bob is a primate. Therefore, Bob > must be human or a monkey, but not both. > > But because of the open world assumption, that's OK. Bob *is* > either a human or a monkey, and not both - but we haven't said > which he is. > Correct. But according to RDF semantics http://www.w3.org/TR/rdf-mt/#sinterp in one interpretation of the above RDF statements Bob may be a human, while in another interpretation Bob may be a monkey. But the RDF semantics doesn't tell us which interpretation to pick -- that's up to the application. In other words, in application1, the above RDF data may be combined with other RDF data (call it r1) that constrains Bob to be a human, while in application2 the above RDF data may be combined with different RDF data (call it r2) that constrains Bob to be a monkey. This does not mean that either application is wrong. It just means that the above RDF data cannot be used with both r1 and r2 *in the same application*. This paper, from last July's International Joint Conference on Artificial Intelligence, discusses this further and explains how a "URI declaration" constrains the possible interpretations corresponding to a URI: http://dbooth.org/2009/denotation/ -- David Booth, Ph.D. Cleveland Clinic (contractor) Opinions expressed herein are those of the author and do not necessarily reflect those of Cleveland Clinic.
Received on Friday, 2 October 2009 18:39:23 UTC