Re: Deduced property from PAUL WARREN on 2014-01-31 (semantic-web@w3.org from January 2014)

From: PAUL WARREN <paul.w.warren@btinternet.com>
Date: Fri, 31 Jan 2014 10:38:36 +0000 (GMT)
To: Aidan Hogan <aidan.hogan@deri.org>, Enrico Franconi <franconi@inf.unibz.it>
Cc: Dave Reynolds <dave.e.reynolds@gmail.com>, "semantic-web@w3.org" <semantic-web@w3.org>
Message-ID: <1391164716.84010.YahooMailNeo@web87402.mail.ir2.yahoo.com>
I think the issue is more fundamental than the use of the word 'inheritance'.  In a recent trial I found that eight out of 12 participants assumed that if a property is transitive, then so is its subproperty.  This was despite the fact that the participants were computer scientists, with some knowledge of ontologies, and that introductory reading before the trial pointed out that this was not the case.

I suggest the problem is with the syllable 'sub', as in subproperty.  Most computer scientists will assume inheritance when they see 'sub' (whether they use the actual word 'inheritance' or not).  This will be the case whether they come from an OO background, or whether they are used to thinking about subclasses in ontologies.  As has been pointed out, properties and their restrictions are inherited by subclasses.  

I am simply saying that this needs to be made more explicit in teaching about ontologies.  It's probably too late now, but a different term from subproperties, e.g. 'property specializations', would be useful.

Cheers,

Paul Warren


________________________________
 From: Aidan Hogan <aidan.hogan@deri.org>
To: Enrico Franconi <franconi@inf.unibz.it> 
Cc: PAUL WARREN <paul.w.warren@btinternet.com>; Dave Reynolds <dave.e.reynolds@gmail.com>; "semantic-web@w3.org" <semantic-web@w3.org> 
Sent: Thursday, 30 January 2014, 18:42
Subject: Re: Deduced property
 

On 30/01/2014 14:55, Enrico Franconi wrote:
>
> On 30 Jan 2014, at 16:29, Aidan Hogan <aidan.hogan@deri.org> wrote:
>
>> Paul, I think Dave's advice is sound: as tempting as it might be, it is *not* helpful to talk about OWL subsumption using phrases like inheritance. This will do more harm than good (esp. since the counter-examples will heavily outweigh the examples).
>
> Not really.
> "Inheritance" (in the object oriented sense) holds true in DL: a property of all the *objects* of a superclass is inherited to all the *objects* of the subclass.
> So, if class C is subsumed by class D, then if all the objects in class D have a property P, then all the objects in class C have the property P.
> This is true in Java, description logics, OWL, Smalltalk, CLOS, etc.

Yes but I think this is quite a jump away from the topic of the thread 
so far. We were discussing the (lack of) "inheritance" of transitivity 
in properties.

Your argument specifically relates to one feature of OWL: 
someValuesFrom. Indeed there's some notion of "inheritance" in an 
object-oriented sense here (as well as for the other property-based 
restrictions on classes). But again this is only a subset of the 
features of OWL (and not the ones we were discussing).

Some "features" of OWL are "inherited". Others are not.

a sCO b. b equivClass c . ⊬ a equivClass c .
a sPO b . b inverseOf c . ⊬ a inverseOf c .
a sPO b . b equivProp c . ⊬ a equivProp c .
a sPO b . b type SymProp . ⊬ a type SymProp .
a sPO b . b type TransProp . ⊬ a type TransProp .
...

... I still maintain my original point that it is *not* helpful to talk 
about OWL subsumption using phrases like inheritance. :)


Cheers,
Aidan

> cheers
> —e.
>
>
>>
>> Hence why the "inheritability" of different OWL features isn't documented (and nor should it be).
>>
>> If you want a non-technical means of introducing the features of OWL, examples using IF -- THEN -- (i.e., rules) will give a sound but incomplete picture. Studying the rules in OWL 2 RL/RDF is a great starting point for anyone wanting to learn a bit about what the *key* entailments of the OWL (2) features are (and without having to get into the formal semantics):
>>
>> http://www.w3.org/TR/owl2-profiles/#Reasoning_in_OWL_2_RL_and_RDF_Graphs_using_Rules
>>
>> The OWL features mean more than what's represented in these rules, but IF you can understand these rules, THEN you'll have a working knowledge of OWL.
>>
>> (Unfortunately though, I feel we're fighting a losing battle with regards the didactic aspects of OWL in the broader sense of it being a *Web standard*. Perhaps the battle is even already lost.)
>>
>> Best,
>> Aidan
>>
>>
>> On 30/01/2014 05:43, PAUL WARREN wrote:
>>> I have come across this problem recently in some work I have been doing
>>> investigating people's understanding of OWL constructs.  You can't
>>> assume that property characteristics are inherited - some are (e.g.
>>> functionality), some aren't (e.g. transitivity and symmetry).  But I
>>> have found no reference in any documentation to this fact.
>>>
>>> Cheers,
>>>
>>> Paul Warren
>>>
>>> ------------------------------------------------------------------------
>>> *From:* Dave Reynolds <dave.e.reynolds@gmail.com>
>>> *To:* semantic-web@w3.org
>>> *Sent:* Wednesday, 29 January 2014, 17:05
>>> *Subject:* Re: Deduced property
>>>
>>> OWL, and the underlying logic, are quite different from object oriented
>>> modelling so using terms like "inheritance" can trip you up. Especially
>>> when it comes to property axioms.
>>>
>>> In the RDF/OWL way of thinking then a property corresponds to set of
>>> pairs of things that are related by the property. So saying
>>>
>>>      :hasParent rdfs:subPropertyOf :hasAncestor
>>>
>>> means, and only means, that the set of pairs of things related by
>>> :hasParent is a subset of the set of pairs of things related by
>>> :hasAncestor.
>>>
>>> It's sets all the way down :)
>>>
>>> Dave
>>>
>>> On 29/01/14 16:47, Jean-Claude Moissinac wrote:
>>>> OK
>>>> I really thought that the transitivity was inherited. I will try to find
>>>> where and how the non-inheritance is specified
>>>> Thank you
>>>>
>>>> --
>>>> Jean-Claude Moissinac
>>>>
>>>>
>>>>
>>>> 2014-01-29 Matthew Horridge <matthew.horridge@stanford.edu
>>> <mailto:matthew.horridge@stanford.edu>
>>>> <mailto:matthew.horridge@stanford.edu
>>> <mailto:matthew.horridge@stanford.edu>>>
>>>>
>>>>     Hi Jean-Claude,
>>>>
>>>>     Asserting
>>>>
>>>>     :hasParent rdfs:subClassOf :hasAncestor
>>>>
>>>>     and
>>>>
>>>>     :hasAncestor rdf:type owl:TransitiveProperty
>>>>
>>>>     does not mean that :hasParent is also transitive.  Transitivity
>>>>     isn’t “inherited” down the property hierarchy, so it’s possible to
>>>>     have a non-transitive sub property of a transitive super property.
>>>>
>>>>     Cheers,
>>>>
>>>>     Matthew
>>>>
>>>>     On 29 Jan 2014, at 08:30, Jean-Claude Moissinac
>>>>     <jean-claude.moissinac@telecom-paristech.fr
>>> <mailto:jean-claude.moissinac@telecom-paristech.fr>
>>>>     <mailto:jean-claude.moissinac@telecom-paristech.fr
>>> <mailto:jean-claude.moissinac@telecom-paristech.fr>>> wrote:
>>>>
>>>>>     No, it's not the answer because hasAncestor is transitive and
>>>>>     hasParent isn't...
>>>>>     (I've a lot of similar situations)
>>>>>
>>>>>     --
>>>>>     Jean-Claude Moissinac
>>>>>
>>>>>
>>>>>
>>>>>     2014-01-29 Richard Cyganiak <richard@cyganiak.de
>>> <mailto:richard@cyganiak.de>
>>>>>     <mailto:richard@cyganiak.de <mailto:richard@cyganiak.de>>>
>>>>>
>>>>>         Jean-Claude,
>>>>>
>>>>>         You’re looking for this (in Turtle syntax):
>>>>>
>>>>>           :hasParent rdfs:subClassOf :hasAncestor.
>>>>>
>>>>>         (Don’t try to read or write RDF/XML directly. You’ll go mad.
>>>>>         Use the friendly syntaxes such as Turtle, or graphical tools.)
>>>>>
>>>>>         Best,
>>>>>         Richard
>>>>>
>>>>>
>>>>>         On 29 Jan 2014, at 16:18, Jean-Claude Moissinac
>>>>>         <jean-claude.moissinac@telecom-paristech.fr
>>> <mailto:jean-claude.moissinac@telecom-paristech.fr>
>>>>>         <mailto:jean-claude.moissinac@telecom-paristech.fr
>>> <mailto:jean-claude.moissinac@telecom-paristech.fr>>> wrote:
>>>>>
>>>>>         > Sorry if my question is very naive, but I'm stuck on this
>>>>>         for a while
>>>>>         > if I go to examples in the document
>>>>> http://www.w3.org/TR/2009/WD-owl2-primer-20090421/
>>>>>         > I just want to add the following axiom (expressed here in my
>>>>>         syntax)
>>>>>         > if
>>>>>         > ?s :hasParent  ?f
>>>>>         > Then
>>>>>         > ?s :hasAncestor ?f
>>>>>         >
>>>>>         > I've checked a lot of documents and I don't figure how to do
>>>>>         it (directly in XML/RDF or interactively with Protégé)
>>>>>         >
>>>>>         > Thank you in adavnace for your help
>>>>>         >
>>>>>         > --
>>>>>         > Jean-Claude
>>>>>         >
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>
>
Received on Friday, 31 January 2014 10:41:56 UTC