- From: <jpan@csd.abdn.ac.uk>
- Date: Thu, 22 Dec 2005 11:08:14 +0000
- To: Alan Rector <rector@cs.man.ac.uk>, Guus Schreiber <schreiber@cs.vu.nl>
- Cc: public-swbp-wg@w3.org
This is a review of the QCR draft at http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/att-0022/QCRs-revised-01.html This is an interesting draft which provides three possible ways to represent some levels of QCRs. I like it very much and I hope the following comments below could help further improve the quality of the draft. Jeff Major Comments ============= * Pattern 2 1. Title of Pattern 2 is not very precise as we use more than just subPropertyOf axioms; e.g., we also use range axioms in this trick. I suggest we adjust the words a bit and use, e.g., 'Work around' using sub-properties as the title of Pattern 2. 2. The example used in the Discussion can be extended with some facts (see below) which could made it clearer why the trick sometimes is not enough. Give the following class axiom Class( Minimal_Italian_Dinner partial intersectionOf( Restriction( has_starter cardinality(1)) Restriction( has_main_course cardinality(1)) Restriction( has_desert cardinality(1)) Restriction( has_course someValuesFrom(IceCream_course) ))), the (subPropertyOf and range) property axioms about has_starter, has_main_course, has_desert and has_course axioms) and the following facts Individual (dinner1 value(has_starter s1) value(has_main_course m1) value(has_desert d1) value(has_course Individual (c1 type(IceCream_course)))) DifferentIndividuals(c1 d1), we have dinner1 as an instance of the Minimal_Italian_Dinner class, even if IceCream_course is a sub-class of Desert. Note that different sub-properties of has_course **can** have the same range; therefore, the above axioms do not entail that dinner has_desert c1. This example indicates that the use of the super property can easily open a back door for unintended individuals. 3. Another side affect of this trick is the introduce of (unnecessary) global range constraints (for the new sub-properties). Note that QCRs are simply some local constraints. * Pattern 3 1. Following the style of the first two patterns, we should introduce the abstract syntax first, then RDF syntax. 2. There might be an issue of the abstract syntax: valuesFrom is not included in the OWL standard (there exist allValuesFrom and someValuesFrom, but no valuesFrom). Therefore, it might not be legal in OWL Full. Minor Comments ============= 1. The title tag needs to be updated. 2. OWL abstract syntax: unbalance brackets, "," after the property URI in a restriction (and there should not be).
Received on Thursday, 22 December 2005 11:10:04 UTC