RE: Detecting cycles in OWL

Hi Enrico,

OK, but then, of course, you should be able to embedd in it all the semantics of RDF(s) (and maybe some of the additional OWL-like constructs that you'd like to have), right? This could be an interesting exercise.

                Currently, our goal is to build upon the RDF semantics, for instance, adding semantics of existential and universal quantifiers, logical implications and quoted graphs (starting from Doerthe’s prior formalization of semantics [1])

And how do you reconcile monotonic negation with the "open world" semantics assumed by RDF(s)? Indeed, monotonicity would be just local, and would not be kept after, say, adding new triples coming from other sites.

                You raise a very good point. The negation as failure will be scoped towards a document at a given point in time; new triples from other sites, or even from the same site at a different time, would not exist within that scope, and hence not result in prior inferences no longer being valid. This closes the world to an extent that is often useful in practice (at least, many use cases I have encountered). As Doerthe mentioned this is not a novel idea and Axel Polleres has done some work in this area (e.g., [2]).

[1] https://biblio.ugent.be/publication/8627430/file/8632611.pdf


[2] https://link.springer.com/content/pdf/10.1007/11762256_26.pdf


W

From: Franconi Enrico <franconi@inf.unibz.it>
Sent: May-05-20 10:24 AM
To: Doerthe Arndt <doerthe.arndt@ugent.be>
Cc: semantic-web@w3.org
Subject: Re: Detecting cycles in OWL

CAUTION: The Sender of this email is not from within Dalhousie.
Hi,

Il giorno 5 mag 2020, alle ore 12:56, Doerthe Arndt <doerthe.arndt@ugent.be<mailto:doerthe.arndt@ugent.be>> ha scritto:
You are right that N3 allows existentials in the head of a rule and that the language is not decidable. It is however still possible to put restrictions on the rules themselves (as it is also done for existential Datalog, there are too many publications about it to add a reference) or to implement different reasoning strategies which are incomplete for some cases but guarantee termination (this is done in EYE [1]).
Scoped negation as failure in N3 is not formalised yet. We  plan to follow exiting publications [2,3] but adjust them to our applications. We currently discuss this adjustment and also the concrete way to express that in N3 (so suggestions are welcome).
The basic idea of scoped negation as failure is that you keep the monotonicity by explicitly mentioning the scope of your negation. So, you have a "closed sub-world" instead of a "closed world".

So, you have in mind one of the variants of datalog± with an added monotonic negation, is it?

OK, but then, of course, you should be able to embedd in it all the semantics of RDF(s) (and maybe some of the additional OWL-like constructs that you'd like to have), right? This could be an interesting exercise.

And how do you reconcile monotonic negation with the "open world" semantics assumed by RDF(s)? Indeed, monotonicity would be just local, and would not be kept after, say, adding new triples coming from other sites.

The spec is not ready yet but we are working on it and hope to be able to release a first version soon. If you are interested, I point you to my paper about N3's semantic [4] where many aspects are already covered (as mentioned above, the negation is still missing).

Thanks, I'll look at that.
cheers
--e.