[Owlim-discussion] How to use "NOT EXISTS" relationship filter in rule
Dear sirs, I need to write a rule, that will be satisfied, if there exists one relationship between individuals and one not. For example: x y [Constraint x != y ] x p y [Constraint p != , x != y] ... --- x y So, when there are two individuals, that are in relationship realestate:sameRealEstateAsCandidate, and there is not relationship owl:differentFrom between them, then x y. But, when I made a test for following triples ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2 ex:individual1 owl:differentFromex:individual2 Trree created following undesired triple. ex:individual1 ex:rel ex:individual2 Could you please help me in this? I really need this inference, it is one key rules of our matching process. When I looked in the example shown in the start of OWL2 RL pie file, there is wery similar example, thus I belevie I am just making some mistake. Note, that both triples in the premise part are implicit triples. with kindest personal regards miro ___ OWLIM-discussion mailing list OWLIM-discussion@ontotext.com http://ontotext.com/mailman/listinfo/owlim-discussion
Re: [Owlim-discussion] How to use "NOT EXISTS" relationship filter in rule
Hello miro, I think the rule you have created is correct, however you are getting in to a tricky area to do with negation. OWLIM's inferencer can only infer new statements when adding explicit statements - in effect, it assumes monotonicity. Added to this, it is an incremental reasoner and will apply all rules to each individual statement as it is added. Therefore the behaviour you are seeing could well happen depending on the order in which statements are inserted. For example, If you assert the statements in this order: 1. ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2 => the inferencer fires the rule to get ex:individual1 ex:rel ex:individual2 2. ex:individual1 owl:differentFromex:individual2 => doesn't cause any rule to fire, but won't remove any statements either In this order: 1. ex:individual1 owl:differentFromex:individual2 => doesn't cause any rule to fire 2. ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2 => still no rule fires Essentially, I believe the problem to be caused by modelling issues within your ontology. If you think about it, "not equal to owl:differentFrom" is already getting to be rather double-negative. It seems that you have a model where some real-estates are the same unless you state that they are different. Is it not possible to organise things so that you make owl:sameAs statements about these entities rather than owl:differentFrom statements? I hope this helps, barry On 24/11/11 14:37, Miroslav Líška wrote: Dear sirs, I need to write a rule, that will be satisfied, if there exists one relationship between individuals and one not. For example: x y [Constraint x != y ] x p y [Constraint p != , x != y] ... --- x y So, when there are two individuals, that are in relationship realestate:sameRealEstateAsCandidate, and there is not relationship owl:differentFrom between them, then x y. But, when I made a test for following triples ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2 ex:individual1 owl:differentFromex:individual2 Trree created following undesired triple. ex:individual1 ex:rel ex:individual2 Could you please help me in this? I really need this inference, it is one key rules of our matching process. When I looked in the example shown in the start of OWL2 RL pie file, there is wery similar example, thus I belevie I am just making some mistake. Note, that both triples in the premise part are implicit triples. with kindest personal regards miro ___ OWLIM-discussion mailing list OWLIM-discussion@ontotext.com http://ontotext.com/mailman/listinfo/owlim-discussion ___ OWLIM-discussion mailing list OWLIM-discussion@ontotext.com http://ontotext.com/mailman/listinfo/owlim-discussion
Re: [Owlim-discussion] How to use "NOT EXISTS" relationship filter in rule
Hello Miro,
the rule:
x y [Constraint x != y ]
x p y [Constraint p !=
, x != y]
...
---
x y
would fire for any relation that links X with Y and is not equal to
owl:differentFrom ...
so one such relation is (x y)
that you've just asserted :-)
Not knowing your model (ontology) and intended use, my advice could look
a bit vague to you ... but you may reject results from such
contradictory inference at query time using optionals, e.g:
select where {
...
?x realestate:sameRealEstateAsCandidate ?y .
optional {
?x realestate:differentRealEstate ?y .
?x ?foo .
}
FILTER (!bound(?foo)) ..
...
}
inthe above quert a bound value to ?foo indicates that all the patterns
within the optional are satisfied so you could skip all the solutions
where ?y is bound to an instance that is NOT to be considered a
sameRealEstate to ?x
HTH,
Damyan Ognyanov
Ontotext AD.
On 12/1/2011 11:52 AM, Barry Bishop wrote:
Hello miro,
I think the rule you have created is correct, however you are getting
in to a tricky area to do with negation. OWLIM's inferencer can only
infer new statements when adding explicit statements - in effect, it
assumes monotonicity. Added to this, it is an incremental reasoner and
will apply all rules to each individual statement as it is added.
Therefore the behaviour you are seeing could well happen depending on
the order in which statements are inserted. For example,
If you assert the statements in this order:
1. ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2
=> the inferencer fires the rule to get ex:individual1 ex:rel
ex:individual2
2. ex:individual1 owl:differentFromex:individual2
=> doesn't cause any rule to fire, but won't remove any statements either
In this order:
1. ex:individual1 owl:differentFromex:individual2
=> doesn't cause any rule to fire
2. ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2
=> still no rule fires
Essentially, I believe the problem to be caused by modelling issues
within your ontology. If you think about it, "not equal to
owl:differentFrom" is already getting to be rather double-negative. It
seems that you have a model where some real-estates are the same
unless you state that they are different.
Is it not possible to organise things so that you make owl:sameAs
statements about these entities rather than owl:differentFrom statements?
I hope this helps,
barry
On 24/11/11 14:37, Miroslav Líška wrote:
Dear sirs,
I need to write a rule, that will be satisfied, if there exists one
relationship between individuals and one not.
For example:
x y [Constraint x != y ]
x p y [Constraint p !=
, x != y]
...
---
x y
So, when there are two individuals, that are in relationship
realestate:sameRealEstateAsCandidate, and there is not relationship
owl:differentFrom between them, then x y.
But, when I made a test for following triples
ex:individual1 realestate:sameRealEstateAsCandidate ex:individual2
ex:individual1 owl:differentFromex:individual2
Trree created following undesired triple.
ex:individual1 ex:rel ex:individual2
Could you please help me in this? I really need this inference, it is
one key rules of our matching process. When I looked in the example
shown in the start of OWL2 RL pie file, there is wery similar
example, thus I belevie I am just making some mistake. Note, that
both triples in the premise part are implicit triples.
with kindest personal regards
miro
___
OWLIM-discussion mailing list
OWLIM-discussion@ontotext.com
http://ontotext.com/mailman/listinfo/owlim-discussion
___
OWLIM-discussion mailing list
OWLIM-discussion@ontotext.com
http://ontotext.com/mailman/listinfo/owlim-discussion
___
OWLIM-discussion mailing list
OWLIM-discussion@ontotext.com
http://ontotext.com/mailman/listinfo/owlim-discussion
