Syntactic Mapping vs Class level

      Greetings,

      Maybe I read too much into the OWL Guide about how "high up" the articulation should occur. This sentence kind of threw me, "We want simple assertions about class membership to have broad and useful implications" (Right out of the OWL Guide Ch.4 middle para. 2). I took this to mean that articulating at the class level is more "broad and useful" than an articulation at the instance level.  I see now after your explanation and my experimentation over the last few days that it can be a bit difficult to do especially if the two classes are similar, but not equivalent for reasons of granularity etc. The tendency could be to misuse the equivalentClass when the classes really may not be equivalent because they do not have precisely the same instances (OWL Guide Ch.4.1 middle para. 2).  In my case the instances were similar, but the granularity differed...  

      Here's my example: 

      I have a class 9mm Automatic Pistol that has instances Beretta 92FS; Browning Highpower; SigSauer P226 and Glock 17.  All specific 9mm pistols much like your specific horse types.  I also have another class, Beretta 92 Series Pistol, in a separate ontology that has instances FS, MFS, DS, D all of which are more specific types of the Beretta 92 Series.  Originally I was trying to declare the 9mm Automatic Pistol Class an equivalentClass of the Beretta 92 Series class.  While both are SIMILAR, they are NOT equivalent because they do not have precisely the same instances.  

      So my solution was to take the instance Beretta 92FS from the Automatic Pistol Class and the FS instance from the Beretta 92 Series Class and articulate at that level:

                              <Auto_Pistol rdf:ID="Beretta 92FS"> 
                                      <owl:sameIndividualAs rdf:resource="&Beretta;FS"/> 
                              (I had the proper ns defined to make this shorthand work) 
                              </Auto_Pistol> 

      Is there a better way?  

      Is it true that if the articulation is done at the instance level, since in most cases this is the most granular level, granularity and consistency problems are avoided?

      Thanks again for you time, I appreciate it very much 

      Marty Hagenston 



      -----Original Message----- 
      From: Drew McDermott [mailto:drew.mcdermott@yale.edu] 
      Sent: Monday, March 31, 2003 11:03 AM 
      To: www-rdf-logic@w3.org 
      Subject: Re: Ontology Mapping 




         Reply-To: <hagenstonmd@sbcglobal.net> 
         From: <hagenstonmd@sbcglobal.net> 
         Date: Wed, 26 Mar 2003 13:21:05 -0800 

         I would welcome any advice/comments on Ontology Mapping.  I am trying to = 
         put together an exemlar that demonstrates how an Articulation Ontology = 
         works. 

         My question is in reference to Chapter 4 in the Web Ontology Language = 
         Guide Version 1.0 dtd 10 Feb 2003 Ontology Mapping. 

         Assuming you have two separate ontologies (O1, O2) both having separate = 
         conceptualizations you wish to adopt.  In order to extend the desired = 
         conceptualizations,  a separate Articulation Ontology is established to = 
         articulate/map the desired conceptualizations from both sources.  In = 
         order for this concept to work would the Articulation Ontology first be = 
         required to import O1 and O2 before declaring that O1 ClassA = 
         equivalentClass  O2 ClassB ?  Or is it a better practice to precede the = 
         mapping statement in the Articulation Ontology with the URL of each = 
         equivalent class and not bother with the import allowing the URL to = 
         point to the location(s) of the equivalent classes? =20 

      You could do it either way.  The first is called "merging" the two 
      ontologies, i.e., producing an ontology that is the union of the two 
      plus further axioms that express the relationships among their symbols. 
      The second one might call "syntax mapping," since it treats the 
      component ontologies as entities to be referred to, not used. 

      The OWL guide 
      (http://www.w3.org/TR/2002/WD-owl-guide-20021104/#OntologyMapping) 
      is talking about the merging approach. 

         Also, according to the reference, it is more efficient to 
         articulate as  "high up" the class hierarchy as possible, meaning 
         once it is stated  that two classes are equivalent, by definition 
         all their members must  satisfactorily belong to both classes. 

      Where exactly does it talk about how high up the articulation is? 

         Would this obviate the  requirement for a similar articulation at 
         the instance level as long as  the user subscribed to the higher 
         up Articulation Ontology?  What harm  would the redundancy cause 
         if mappings at both levels occurred? 

      I don't see how it would obviate lower-level matchings.  If ontology 1 
      lists a bunch of horses, and ontology 2 lists a bunch of equines, 
      merely knowing that "horse" and "equine" are the same class doesn't 
      tell you which horse in list 1 is the same as which horse in list 2 
      (if any in fact are the same). 

      -- 
                                                   -- Drew McDermott