Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[protege-discussion] SWRL Disjonction

Biswanath Dutta dutta2005 at gmail.com
Fri Apr 3 15:35:07 PDT 2009


Hi,

When your model is defined explicitly and your requirements can be meet by
just running the queries on the Asserted knowledge (as exactly, Protege
SPARQL tab does), then yes, you no need to write any rules (e.g., SWRL, N3)
on top of your ontology. But when you need to infer some new knowledge from
the existing knowledge, then, if you have to have the rules on top of your
ontology. For example,

hasMother(?x, ?y)  ^  hasBrother(?y, ?z) → hasUncle(?x, ?z)

Regarding your second question, yes, sometime it's become difficult to
decide whether something to be considered as a Class or  an Instance. But
again it's not that difficult to decide. Decision comes from your
requirement or purpose in desigining the model. What kind of knowledge you
are going to infer from your knowledge-base. After few exercises in
modelling, it becomes easy to judge that.

Hope this will help you.

With regards,
Biswanath Dutta (bisu)










On Thu, Apr 2, 2009 at 10:59 PM, narjess touzani <narjess_tz at yahoo.fr>wrote:

> I use SWRL rules in oder to infer knwoledge. I have to model pedagogical
> knowledge, then infer them to provide assistance.
> So, if i had understand right, there is no need to write SWRL rules? and
> only from resonnig ontology i can extract whatever information I want?
> Anotehr question, I want to know when must I use class or instance? how can
> I decide to use the one or the other? Because in my example, I used
> "case_study" and "analysis" as instances of the class "Activity" , and
> "Production" instance of the class "activity_type". In your mail you propose
> me a different hierarchy:
>    Activity
>       Production
>          case_study
>          analysis
> where all of them are classes. What is the difference?
> I thought that in an ontology, i must specify class instances. furthermore,
> by specifying instances I will be able to write SWRL rules, that is?
> Best regards
>
>
>
> --- En date de : *Jeu 2.4.09, Thomas Russ <tar at ISI.EDU>* a écrit :
>
>
> De: Thomas Russ <tar at ISI.EDU>
> Objet: Re: [protege-discussion] SWRL Disjonction
> À: "User support for Core Protege and the Protege-Frames editor" <
> protege-discussion at lists.stanford.edu>
> Date: Jeudi 2 Avril 2009, 18h21
>
>
> On Apr 2, 2009, at 1:46 AM, narjess touzani wrote:
>
> > Dear Mr.Thomas,
> > Thank you for your response.
> > I am modelling pedagogical knowledge. And I want to express and
> > infer rules on these knowledge. So I want to express rules like :
> > if activity_type is "production" then activity can be a "case study"
> > one OR an "Analysis" one. If we write this rule in SWRL:
> > activity_type(Production) --> Activity(case_Study) OR
> > Activity(analysis)
> >  So, in this case I need the disjonction operator.
>
> OK.  So how does it help you in further reasoning?
>
> There isn't anything you can conclude based on the results of this
> rule.  So what do you want this to be able to tell you?  If you asked
> if the activity is a "case study", any logical reasoner would have to
> return unknown (or even worse false if you used a closed world
> reasoning system).
>
> Now, if what you are trying to do is narrow things down a bit, then
> you either need to make assertions about the activity being in the
> complement of those types which it is now known NOT to be.
> Alternately, you would need to introduce another class which is the
> union of the case_study and analysis classes, and write your rule to
> assert membership in that class:
>
>     Analytical_Activity <=>  unionOf case_study analysis
>
> By introducing this disjunctive parent of case_study and analysis, you
> can introduce a controlled measure of disjunction, but in a positive
> way that allows you to use the information in further reasoning.
>
> It also seems a bit odd to be using a relation for noting type
> information.  In Frames or OWL, it would be more natural to model
> types as classes and have the built-in class and type reasoning
> mechanisms available to you.  In fact, from your example rule, I'm
> wondering why you don't just build a class hierarchy like:
>
>    Activity
>       Production
>          case_study
>          analysis
>
> and not even bother with SWRL rules at all.
>
>
>
> _______________________________________________
> protege-discussion mailing list
> protege-discussion at lists.stanford.edu<http://fr.mc265.mail.yahoo.com/mc/compose?to=protege-discussion@lists.stanford.edu>
> https://mailman.stanford.edu/mailman/listinfo/protege-discussion
>
> Instructions for unsubscribing:
> http://protege.stanford.edu/doc/faq.html#01a.03
>
>
>
> _______________________________________________
> protege-discussion mailing list
> protege-discussion at lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/protege-discussion
>
> Instructions for unsubscribing:
> http://protege.stanford.edu/doc/faq.html#01a.03
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.stanford.edu/pipermail/protege-discussion/attachments/20090403/26e114c0/attachment.html>


More information about the protege-discussion mailing list