- From: Deborah L. McGuinness <dlm@ksl.Stanford.EDU>
- Date: Tue, 20 Apr 2004 15:43:27 -0700
- To: ewallace@cme.nist.gov
- Cc: public-swbp-wg@w3.org
concerning the notion of qualified cardinalities and role hierachies and restrictions (whether they are local with value restrictions or global with range restrictions), it is true that you can end up modeling information concerning qualified cardinalities using role hierarchies, restrictions, and regular cardinality restrictions, but it does get ugly. for example, consider the case where i want to have at least 2 siblings who are doctors on some new class. i can create a role hasSibling and create a subrole of hasSibling (on which i will put a value or range restriction). call the subrole hasSiblingDoc. then in my newclass, i can have a DOCTOR value restriction on hasSiblingDoc (or i could have made the global range restriction on hasSiblingDoc be DOCTOR) and i can put an atmost 2 cardinality restriction on hasSiblingDoc on the newclass. I could similarly do the same kind of modeling for at least 2 siblings who are doctors on another newclass2 The ugliness arises because i have to make the new subroles with the new either local value restrictions or global range restrictions. also, from a computational perspective - the reasoner still has to do the right counting work at the superrole. for example, if i had another subrole called hasSiblingLawyer and had another newclass - newclass3 that had at least 2 hasSiblingLawyer and also had atleast 2 hasSiblingDoctor, then a reasoner should be able to determine some cardinality restriction information on the hasSibling property of newclass3. clearly it is 2 but if one knows anything about the disjointness of the value restrictions or range restrictions (for example, if i knew that doctor and lawyer were disjoint) then i could determine that the atleast restriction was 4. also, if i know something about uniqueness of known fillers or if i know some particular filler is NOT an instance of the class in the range restriction, then i can also increase the min cardinality. for example, if i know that Joe is a filler of the property hasSiblingDoctor and I know that Joe is NOT a lawyer, then I know that Joe must be distinct from the 2 fillers of the hasSiblingLawyer property and I know that the minimum cardinality of hasSibling on newClass3 must be at least 3. deborah ewallace@cme.nist.gov wrote: >I think Alan meant to send this to the SWBPD list. > >In response to his comment. UML Associations are typically binary and >scoped to the classes at both end-points (so in this sense they are innately >qualified). However, Association names must be unique within a Package. It would >take an intelligent mapping to aggregate semantically similar Associations together >into OWL ObjectProperties, although I agree that this is the "right way" to do it. >The mapping rules that I have seen recently in fact do the opposite, they even >mangle DatatypeProperty names for corresponding UML attributes by pre-pending them >with a UML class name. > >Regarding work arounds for qualified cardinality: Cannot one get the >expressiveness of Qualified Cardinality by using subProperties of a >common property and placing local restrictions on the cardinality of those? >Is that ugly or problematic? > >-Evan > > >----- Begin Included Message ----- > > -- Deborah L. McGuinness Associate Director Knowledge Systems Laboratory Gates Computer Science Building, 2A Room 241 Stanford University, Stanford, CA 94305-9020 email: dlm@ksl.stanford.edu URL: http://ksl.stanford.edu/people/dlm (voice) 650 723 9770 (stanford fax) 650 725 5850 (computer fax) 801 705 0941
Received on Tuesday, 20 April 2004 18:43:26 UTC