- From: Pierre-Antoine Champin <swlists-040405@champin.net>
- Date: Fri, 15 Jun 2007 10:25:34 +0200
- To: Ian Davis <lists@iandavis.com>
- Cc: Pierre-Antoine Champin <swlists-040405@champin.net>, semantic-web@w3.org
Ian Davis a écrit : > > On 14/06/2007 16:11, Pierre-Antoine Champin wrote: >> Consider that I work for two different companies (in the morning and in >> the afternoon). >> Both have a URI for me. Company 1 would state >> >> comp1:champin emp:name "Champin" ; >> emp:salary "1000€" . >> >> Company 2, on the other hand, would state >> >> comp2:champin emp:name "Champin" ; >> emp:salary "2000€" . >> >> using the same standardized properties, which happen to be functional. >> >> It would seem legitimate to state that >> >> comp1:champin owl:sameAs comp2:champin . >> >> But that would lead to inconsistency (two different values for a >> functional property). >> >> Both URIs denote me, but not the same "me", only the "me" I am from the >> point of view of each company. > > You modelled it such that each person may have only a single salary, yet > you found someone with two. Either the data is wrong or your model is > wrong. That's all the inconsistency tells you. ok, but which one is wrong ? * the ontology designer, by overconfidently making emp:salary functional? * the companies, by overconfidently using that property (note that in the context of each company, the property *is* actually fonctional)? * myself, by overconfidently trying to merge those unreconcilable data (but damn it, they both are about "me") * myself in overconfidently stating that both company are talking about the *exact* same entity my argument is that the "error" is at the last point, while you consider it is at some higher point. Note that the higher the "error", the harder it will be to fix it. So I think my position is more scalable. But of course, we need a robust way to relate comp1:champin to comp2:champin... pa
Received on Friday, 15 June 2007 08:25:44 UTC