[protege-owl] How to use subproperties?

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

[protege-owl] How to use subproperties?

Gerhard Austaller-4
Hi

Could somebody give me an example where it makes sense to create
subproperties and what benefit  could gained from it? E.g. if I have a
property hasSibling, I could further create subproperties hasBrother
and hasSister. Would that make sense? If so, why?

Further, how can I decide if a candiate for a subproperty is really a
proper subproperty? With subclasses I always can ask myself, if the
subclass "is-also-a" superclass but what's the question to ask myself
with subproperties?

Thanks
Gerhard

_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03 
Reply | Threaded
Open this post in threaded view
|

Re: [protege-owl] How to use subproperties?

Thomas Russ

On Nov 28, 2006, at 12:03 AM, Gerhard Austaller wrote:

> Hi
>
> Could somebody give me an example where it makes sense to create
> subproperties and what benefit  could gained from it? E.g. if I have a
> property hasSibling, I could further create subproperties hasBrother
> and hasSister. Would that make sense? If so, why?

Because it gives you more flexibility in querying as well as asserting.
On the query side, you can use more general relationships to get  
information, in other words you might want to get all siblings and  
use the superproperty to get that.

On the more specific side, a proper definition of the subproperty  
with appropriate range restrictions gives you additional inferences.  
The filler of hasBrother would be a MalePerson, which is information  
that you would not directly have with just a hasSibling property.  
You would have to assert that separately.

>
> Further, how can I decide if a candiate for a subproperty is really a
> proper subproperty? With subclasses I always can ask myself, if the
> subclass "is-also-a" superclass but what's the question to ask myself
> with subproperties?

Pretty much the same question.  If the subproperty is a narrower,  
more specialized version of the superproperty, then it is correctly  
defined as a subproperty.  This can be thought of in terms of logical  
implication.  "If R-sub always implies R, then it is likely to be a  
subproperty."  In your example, hasBrother always implies hasSibling,  
but not vice versa.

If it helps, you can think about this in terms of sets.  A property  
defines a set of binary tuples.  If another property defines a set of  
binary tuples that is always a subset of the first set, then you have  
a subproperty.  This is true of the sets <x,y: brother(x,y)> is a  
subset of <x,y: sibling(x,y)>.



>
> Thanks
> Gerhard
>
> _______________________________________________
> protege-owl mailing list
> [hidden email]
> https://mailman.stanford.edu/mailman/listinfo/protege-owl
>
> Instructions for unsubscribing: http://protege.stanford.edu/doc/ 
> faq.html#01a.03

_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03 
Reply | Threaded
Open this post in threaded view
|

Re: [protege-owl] How to use subproperties?

Kaarel Kaljurand
In reply to this post by Gerhard Austaller-4
Hi,

On 11/28/06, Gerhard Austaller <[hidden email]> wrote:
>
> Further, how can I decide if a candiate for a subproperty is really a
> proper subproperty? With subclasses I always can ask myself, if the
> subclass "is-also-a" superclass but what's the question to ask myself
> with subproperties?
>

E.g. form the following sentence:

Everybody who loves something also likes it.

If you think that this sentence holds then "love" is a sub-property of "like"...
You can use this sentence template with any properties given that they can
be seen as English (transitive) verbs.

--
kaarel
_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03 
Reply | Threaded
Open this post in threaded view
|

[protege-owl] Why are equivalent classes graphically subsumed when creating a taxonomy?

Alan March
When Protégé is instructed to assemble a taxonomy via a resoner, classes
detected as equivalent are placed in a way I find somehow confusing.
Equivalent classes are placed as siblings in a tree, and for each of them,
its equivalent(s) is(are) placed, in gray, alongside the "main" one. Also,
each equivalent class is placed as a child node which itself has a child
node which repeats the parent node with the comment "recursive" in enclosed
in brackets.

Mi question is: is this necessary? I am not quite clear as to the reason why
equivalent classes should appear as graphically subsumed in a tree, as this
strikes me as rather counterintutive. Wouldn't it be more appropiate to just
create one node with equivalent classes placed alongside each other, with no
subsumptions? Should there be some theoretical reason not to do this,
couldn't an option be created by which the user may select the manner in
which to create the tree?

Or am I missing something?

Best regards,

Alan

_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03 
Reply | Threaded
Open this post in threaded view
|

Re: [protege-owl] Why are equivalent classes graphically subsumed when creating a taxonomy?

Boontawee Suntisrivaraporn
On 11/29/06, Alan March <[hidden email]> wrote:

> When Protégé is instructed to assemble a taxonomy via a resoner, classes
> detected as equivalent are placed in a way I find somehow confusing.
> Equivalent classes are placed as siblings in a tree, and for each of them,
> its equivalent(s) is(are) placed, in gray, alongside the "main" one. Also,
> each equivalent class is placed as a child node which itself has a child
> node which repeats the parent node with the comment "recursive" in enclosed
> in brackets.
>
> Mi question is: is this necessary? I am not quite clear as to the reason why
> equivalent classes should appear as graphically subsumed in a tree, as this
> strikes me as rather counterintutive. Wouldn't it be more appropiate to just
> create one node with equivalent classes placed alongside each other, with no
> subsumptions? Should there be some theoretical reason not to do this,
> couldn't an option be created by which the user may select the manner in
> which to create the tree?

Hi Alan,

I don't think that is necessary, and I completely agree with you that
Protege should represent all classes that are inferred equivalent as a
single node in the taxonomy. It can either pick one as a
representative of the node with the rest appear in braces or use all
of them to represent the node.

I think the problem is that Protege inquires the reasoner for the
parent classes of all classes, without being aware of among them there
may be equivalent ones. As a solution, there are two possibilities:
(i) Protege can check the parent sets of sibling classes whether they
contain each other, i.e. A in Parents(B) and B is Parents(A). If so,
they are equivalent. (ii) Protege can inquire the reasoner for all
equivalent classes.

Cheers,
Meng
_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03 
Reply | Threaded
Open this post in threaded view
|

Re: [protege-owl] Why are equivalent classes graphically subsumed when creating a taxonomy?

Michael Wessel-2
In reply to this post by Alan March
Am Wednesday 29 November 2006 13:07 schrieb Alan March:

> Mi question is: is this necessary? I am not quite clear as to the
> reason why equivalent classes should appear as graphically subsumed
> in a tree, as this strikes me as rather counterintutive. Wouldn't
> it be more appropiate to just create one node with equivalent
> classes placed alongside each other, with no subsumptions? Should
> there be some theoretical reason not to do this, couldn't an option
> be created by which the user may select the manner in which to
> create the tree?
> Or am I missing something?

Sounds perfectly reasonable to me. This is how RacerPorter displays
the taxonomy.

Regards,

Michael
_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03 
Reply | Threaded
Open this post in threaded view
|

Re: [protege-owl] Why are equivalent classes graphically subsumed when creating a taxonomy?

Alan March
In reply to this post by Boontawee Suntisrivaraporn
> I think the problem is that Protege inquires the reasoner for
> the parent classes of all classes, without being aware of
> among them there may be equivalent ones. As a solution, there
> are two possibilities:
> (i) Protege can check the parent sets of sibling classes
> whether they contain each other, i.e. A in Parents(B) and B
> is Parents(A). If so, they are equivalent. (ii) Protege can
> inquire the reasoner for all equivalent classes.
>
> Cheers,
> Meng

Yes. I think the reason why Protégé represents this situation as it does is
probably related to what you mention. The trick here, I believe, probably
relates to separating the GUI from the "business layer" (to put it in
"corporate" terms)

_______________________________________________
protege-owl mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-owl

Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03