Re: [OEP] Other QCR approaches

Thanks for noticing the typos ... comments:

At 16:45 -0400 10-06-2005, Christopher Welty wrote:
>public-swbp-wg-request@w3.org wrote on 06/09/2005 02:57:15 PM:
>>  4) decouple a DAML+OIL-like typed QCR into allValuesFrom + OWL QCR. This
>is not
>>  applicable to any of Alan's use cases btw, but only in simpler cases.
>For example:
>>
>>  A typical date (as a meeting btw two prospective lovers) has exactly two
>participants ->
>>
>  > Class(TypicalDate
>>      subClassOf(Restriction(
>>        hasParticipant) cardinality(2)))
>>      subClassOf(Restriction(
>>        hasFinger allValuesFrom(Person))))
>
>Assuming you mean "hasParticipant" instead of "hasFinger" (which would be
>a very atypical date indeed), I have no idea what you are claiming here.
>This is just a regular allvaluesfrom restriction.  QCRs come into play
>when you want to say something like a "typical date has exactly two
>participants and exactly one is male and exactly one is female".  Your
>example, at least, does not demonstrate QCRs at all.

AFAIK, a QCR is a cardinality constraint with a type constraint, then 
this is a way of representing QCRs when a universal restriction is 
applicable. Being in a hurry, as I'd said, I've introduced two 
copy&paste typos: the first is here, I meant:

>  Class(TypicalDate
>      subClassOf(Restriction(
>        hasParticipant) cardinality(2)))
>      subClassOf(Restriction(
>        hasParticipant allValuesFrom(Person))))

Indeed a strange date with fingers (stop stop here!).
The second typo is in the next approach (see below).

>  > I also apply sometimes a reified approach:
>>
>>  5a) reify cardinality as a property, for example:
>>
>>  DatatypeProperty(reifiedCardinality
>>     range(xsd:int))
>>
>>  Class(NormalHand
>>      subClassOf(Restriction(
>>          hasPart someValuesFrom (intersectionOf
>>           Finger
>>           Restriction(
>>            reifiedCardinality oneOf(5)))))
>
>do you mean oneOf(4)?

No, I mean oneOf(5), unless we are talking of Disney's ducks

>  >     subClassOf(Restriction(
>>          hasPart someValuesFrom (intersectionOf
>>           Thumb
>>           Restriction(
>>            reifiedCardinality oneOf(5))))))
>
>do you mean oneOf(1)?

Of course, and this is the other typo I talked about

>Why would you do this?  You get nothing from it other than a way to "hide"
>the QCR from OWL completely.  You would have to build your own inference
>engine that knew something about this kind of cardinality.  You can't
>write axioms in OWL to give you the behavior you want, and you couldn't
>conclude with a standard OWL reasoner, for example, that a hand with four
>fingers and a thumb is a normalHand.

You cannot conclude much anyway, Chris: no real QCR is possible in 
OWL; some workarounds, like creating dummy subproperties or using the 
universal+cardinality when applicable, do work, but in all other 
cases, we use modelling approximations.
Reification is one way of approximating: when one reifies a predicate 
into an individual looses the expressive and reasoning power 
available for predicates; similarly, when one reifies a 
quantification into a restriction, one looses corresponding powers. 
One eventually gains a different modelling style and can apply 
meta-level axioms.

>  > 5b) reify cardinality as a property, and reify the Q(C)R as a class, for
>example:
>>
>>  Class(NormalHandedness
>>      subClassOf(Restriction(
>>        settingFor someValuesFrom(intersectionOf
>>         Hand
>>         Restriction(
>>          hasPart someValuesFrom(Finger)
>>         Restriction(
>>          hasPart someValuesFrom(Thumb))
>>      subClassOf(Restriction(
>>        settingFor someValuesFrom(intersectionOf
>>         Finger
>>         Restriction(
>>          reifiedCardinality oneOf(5)))))
>>      subClassOf(Restriction(
>>        settingFor someValuesFrom(intersectionOf
>>         Thumb
>>         Restriction(
>>          reifiedCardinality oneOf(5)))))
>>      subClassOf(Restriction(
>>        settingFor cardinality(6))))
>
>This is totally confusing.  If you really think you have a valid alternate
>approach, can you verify that your example conveys what you intend, where
>did the "6" come from???? What is "settingFor" supposed to denote?
>"Handedness" in (at least US) English refers to which hand is dominant
>(i.e. "right handed" or "left handed").  I assume that is NOT what you
>mean here, but for the life of me I can't tell what you DO mean.

Besides terminological issues (I can call it "NormalHand"), this is a 
way of providing a more "invasive" reification: NormalHand would be 
here a reified relation that holds between a hand and five fingers, 
one of which is a thumb. the "settingFor" property allows to 
explicitly type the link between the reified relation and its 
arguments (therefore, the total amount of arguments is 6: hand, and 
five fingers). For the rest, it applies the approach (1) and the 
approach (5a).

I see these last patterns require more detailed explanation and use 
cases, and I'll provide it on email, and on the next OEP telecon 
(hopely less depopulated ;))))

Ciao
Aldo

>-Chris
>
>
>>
>>  Approach (5b) uses a pattern similar to the approach 2 from the n-ary
>relations
>>  note, but it also reifies cardinality restrictions as in (5a).
>>  In general, I notice that (5b) is more precise than (5a), because it
>separately
>>  states the actual exact cardinality for this definition of normal
>handedness.
>>  Moreover, other assertions on hands can be made without them impacting
>on the
>>  definition of normal handedness.
>>  The approach in (5b) comes from a more general pattern that can be
>applied to other
>>  parametric constraints, like time-indexed properties, and many other
>applications.
>>  The exemplification with time-indexed properties could be part of a
>dedicated note,
>>  which I'm proposing in a separate message.
>>
>>  Sorry for this quick and probably messed-up explanation, but I wanted to
>submit it
>>  before the OEP telecon (in five minutes ...).
>>  Cheers
>>  Aldo
>>  --
>
>>
>>
>>
>>  Aldo Gangemi
>>  Research Scientist
>>  Laboratory for Applied Ontology
>>  Institute for Cognitive Sciences and Technology
>>  National Research Council (ISTC-CNR)
>>  Via Nomentana 56, 00161, Roma, Italy
>>  Tel: +390644161535
>>  Fax: +390644161513
>>  also.gangemi@istc.cnr.it
>>  http://www.istc.cnr.it/createhtml.php?nbr=71


-- 



Aldo Gangemi
Research Scientist
Laboratory for Applied Ontology
Institute for Cognitive Sciences and Technology
National Research Council (ISTC-CNR)
Via Nomentana 56, 00161, Roma, Italy
Tel: +390644161535
Fax: +390644161513
aldo.gangemi@istc.cnr.it
http://www.istc.cnr.it/createhtml.php?nbr=71

Received on Friday, 10 June 2005 22:21:57 UTC