- From: Guus Schreiber <schreiber@cs.vu.nl>
- Date: Thu, 14 Apr 2005 14:36:49 +0200
- To: Natasha Noy <noy@smi.stanford.edu>
- CC: public-swbp-wg@w3.org
Natasha, A few points about the instance pattern: [[ 1. Define a property hasPart. ]] Earlier in the document we suggest to prefer partOf over hasPart, so I suggest we use partOf here, unless there is a good reason to use hasPart (and then we should mention this explicitly). [[ 2. If we are using OWL (and not RDF Schema), ...... ]] If you use OWL, you're also using RDF Schema (either completely (Full) or with restrictions (DL)). I suggest to delete the text between the parentheses. [[ 3 Define domain and range for the property hasPart (i.e., a class Item) ]] I've become convinced that domain and range should be almost obsolete in ontologies, It restricts reuse of c.q. mapping to a similar property in another ontology. I prefer using the OWL local range restrictions, as their ontological commitment is a lot less. In this case it would mean defining two OWL allValuesFrom restrictions on class Item: one for the partOf and one for the hasPart property. Of course, this approach is more complicated for people to use and it also requires the use of OWL. We could mark this as a choice people can make: the might want to use domain and range, but should be aware of the consequences (and that there is an alternative). Also, I think the name of the class Item is too genera, the name might suggest to the ontology user that "Item" is more general than intended. I would prefer CarPart here, which seems to be the natural term. It assumes the partOf property is reflexive, but we may skip over this issue and refer to later in the document. Guus -- Free University Amsterdam, Computer Science De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands Tel: +31 20 598 7739/7718; E-mail: schreiber@cs.vu.nl Home page: http://www.cs.vu.nl/~guus/
Received on Thursday, 14 April 2005 12:36:53 UTC