No subject

good reasons for the current Higgins and Protege top-level models
to be as they currently are. However, I currently do not see the 
need of any element of the Higgins top-level model (the one at
http://www.eclipse.org/higgins/ontologies/2006/higgins.owl or
http://www.webkb.org/kb/nit/physEntity/animate/higgins_top_level.html)
for representing information about a person in a way that eases
knowledge sharing/re-use, even at a low level (e.g., via RDF+OWL). 

For instance, I currently do not understand why 35 XSD classes
have been "duplicated" into 35 higgins classes and 35 higgins 
properties; one example, higgins#float, which has for signature
(in the FL notation):  higgins#float (higgins#Float, xsd#float).
Why not directly using the 35 XSD classes?

Let's also take the instance example used at the end of 
http://www.eclipse.org/higgins/ontologies/2006/person-with-friend.owl
which I translated in FCG at the end of
http://www.webkb.org/kb/nit/physEntity/animate/higgins_person.html

1) A direct translation in FCG would have been:
[a pwa#Person ,
  pwa#firstname: (a higgins#NormalizedStringSimpleAttribute,
                     higgins:normalizedStringSimpleValue: 
                           the xsd#normalizedString "Mary"),
  pwa#surname: (a higgins#NormalizedStringSimpleAttribute,
                      higgins:normalizedStringSimpleValue: 
                           the xsd#normalizedString "Ruddy"),
  pwa#postalAddress: (a pwa#PostalAddress,
                         pwa#country: the higgins#String "USA",
                         pwa#city: the higgins#String "Provo",
                         pwa#postalCode: the higgins#String "12345",
                         pwa#streetAddress: 
                                 the higgins#String "123 Main St."),
  pwf#friend:
    (a higgins#SubjectRelationship,
         higgins#contextURI: the xsd#anyURI
              "http://example.com/robertjones/public-business-card",
         higgins#subjectCUID: "bob") ];


2) The above representation is very low-level and unlikely to be 
"comparable" with other person-related representations that are not 
using the Higgins top-level model. In my actual translation I got
rid of the added complexity related to strings and I assume it was 
ok to declare pwa#Person as a subtype of higgins#SubjectRelationship
(which I did in the ontology):
[a pwa#Person,
  pwa#firstname: "Mary",  pwa#surname: "Ruddy",
  pwa#postalAddress: (a pwa#PostalAddress ,  pwa#country: "USA" ,
                       pwa#city: "Provo" ,  pwa#postalCode: "12345",
                       pwa#streetAddress: "123 Main St."),
  pwf#friend:
   (a pwa#Person,
       higgins#contextURI:
        "http://example.com/robertjones/public-business-card",
       higgins#subjectCUID: "bob") ];

3) This last representation is still far from ideal for knowledge
sharing and re-use. A better representation would be:
[a pwa#Person,  pwa#firstname: "Mary" ,  pwa#surname: "Ruddy",
  pwa#postalAddress: (a pwa#PostalAddress, pwa#city: pm#Provo,
                           pwa#streetAddress: "123 Main St."),
  pwf#friend:
   (a pwa#Person "http://example.com/robertjones/public-business-card",
        dl#name: "bob") ];
In this representation, the second person (the friend) and the city
have a unique identifier (hence, no need to specify the country and
the postal code anymore, they are associated to pm#Provo which may
be declared as an instance of wn#city for example (see
http://www.webkb.org/bin/categSearch.cgi?explanations=&categ=%23city).

4) More generally, using only types from WordNet (you can check their
supertypes by using the previous URL and replacing "city" by any of 
the other type names below), an even better representation is:
[a #person,  #first_name: "Mary",  #surname: "Ruddy",
  #mailing_address: (a #mailing_address, #city: pm#Provo,
                        #street_address: "123 Main St."),
  #friend: (a #person "http://example.com/robertjones/public-business-card",
                #name: "bob") ];
This representation is good for knowledge entering and sharing because
- the structure is simple and the type names are intuitive,
- the types belong to a large ontology (a lot of what can be said 
  about people can be represented without having to declare new types;
  for example, see http://www.webkb.org/kb/persons.html),
- new types can be added to the shared ontology by anyone (via WebKB-2
  or any other knowledge server allowing updates on a similar ontology),
- the simple structure and shared ontology increase the chances for
  people-related representations (created by different persons/agents)
  to be comparable and hence to be retrieved and exploited for 
  inferencing purposes.
With this approach, no element of the Higgins top-level model is needed.


Philippe