RE: RDF and OWL rules

> I ended up re-ordering by owl rules to match your order and gave them
> the same names as yours (my rdfs rules were already in the same order -
> I also gave them the same names also).

ok

> I seem to have mostly a subset of your rules - i.e. the rules I have
> (with the exception of a couple) correspond to rules that you have. I
> don't have a number that you have:
>
> rule3r2
> rule5e2, rule5e2

guess you meant e1?

> rule7p1, rule7p2, rule7p3
> rule8s1, rule8s2, rule8s3, rule8s6
> rule9o1, rule9i2, rule9u1, rule9f2, rule9i2
> rule10c1, rule10c2, rule10c3, rule10c4, rule10r1, rule10a1
>
> Rules that you don't seem to have (but may have been derivable by other
> means?):
>
> someValuesFrom, owl:equivalentClass, owl:complementOf

right for owl:someValuesFrom
for owl:equivalentClass we have
  owl:equivalentClass rdfs:subPropertyOf rdfs:subClassOf;
                      a owl:SymmetricProperty;
                      a owl:TransitiveProperty.
  {:rule7c1. ?a rdfs:subClassOf ?b. ?b rdfs:subClassOf ?a}
                      => {?a owl:equivalentClass ?b}.

for owl:complementOf we are still trying negation (by declaration)
  owl:complementOf a owl:SymmetricProperty.
  {:rule60n1. ?A owl:complementOf ?B. ?C a ?A} => {?C neg:type ?B}.
  {:rule60n2. ?A owl:oneOf ?L. ?L neg:item ?C} => {?C neg:type ?A}.
  {:rule60n3. ?A owl:complementOf ?B. ?C neg:type ?A} => {?C a ?B}.


> A few notes:
>
> - seems like either rule4p1 or rule4p2 is necessary, but not both
> (assuming equivalentProperty is stated as symmetric)

right, good catch; corrected

> - rule8s6 seems unnecessary (derivable by other rules)

Well, if I leave it out we get 5 more tescases with
"no proof found" (but we test with --step 100000)
I believe that some redundacy is not bad in practice?


> - what are the semantics of math:proofCount? Are your cardinality
> inconsistency rules making a unique names assumption?

math:proofcount is an experimental builtin
to count how many different proofs can be
found for the subject graph posed as query
against the loaded kb (just to experiment)
(will check no-una again but I thought that
http://www.w3.org/2002/03owlt/maxcardinality/inconsistent001
http://www.w3.org/2002/03owlt/maxcardinality/inconsistent002
were done OK as they needed explicit owl:differentIndividualFrom
(now called owl:differentFrom) statements)


> - Seems that most inconsistencies (with exception of datatype violations
> and some cardinality issues) stem from something being both the sameAs
> and differentFrom something else or a class being both a subClassOf and
> disjointWith/complementOf another class. How does neg:type fit in?

we have to work that out...
good idea to think about that one

> Looks promising - the commonality of results created from separate
> efforts is a good sign. I'm going to bring my rules up to speed with
> yours before publishing them. Then I'll compare my test results with
> yours and try to tackle some of the failing cases.

looks very promising

-- ,
Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/

Received on Tuesday, 1 April 2003 15:24:28 UTC