Re: OWL 2.0 --> Rules

Hi Denny --

This may put the cat among the OWLs (:-), but here goes....

In your note, below, you put forward 9 requirements, and your 10th one is 
"not to forget that we have to integrate with rules later".

My question is, why later?  Why not now?

The "Semantic Web Layer Cake" has Rule Languages next to OWL at the same level.

There's a rule language live online [1,2,3] that appears to meet most (and 
perhaps all) of your requirements.

Or have I missed the point somehow?

                         Cheers,  -- Adrian

[1]  http://www.w3.org/2004/12/rules-ws/paper/19/

[2] 
http://www.reengineeringllc.com/demo_agents/FeaReferenceModelOntology1.agent

[3]  http://www.reengineeringllc.com/demo_agents/RDFQueryLangComparison1.agent




At 05:50 PM 5/31/2005 +0200, you wrote:

>Hello all,
>
>quite some while ago the question of OWL 2.0 was rised here, and I wrote 
>already two long replies with a wishlist - but both were never sent and 
>got lost in digital nirvana, one due to a hardware, the second due to a 
>software failure. Well, let's hope this one passes finally through. That's 
>why this answer is so late.
>
>Sorry for the lengthy Mail. But I tried to structure it a bit and make it 
>readable, so I hope you find some interesting stuff here. So, here is my 
>wishlist.
>
>1) I would like yet another OWL language, call it OWL RDF or OWL 
>Superlite, or whatever. This is like the subset of OWL Lite and RDFS. For 
>this the difference between of owl:Class and rdf:Class needs to be somehow 
>standardly solved. Why is this good? It makes moving from RDF to OWL 
>easier, as it forces you to keep Individuals, Classes and Relations in 
>different worlds, and forgets about some of the more sophisticated 
>constructs of RDF(S) like lists, bags and such. This is a real beginners 
>language, really easy to learn and implement.
>
>2) Defined Semantics for OWL FUll. It is unclear -- at least to me -- what 
>some combinations of RDF(S)-Constructs and OWL DL-constructs are meant to mean.
>
>3) Add easy reification to OWL. I know, I know, making statements about 
>statements is meant to be the root of all evil, but I find it pretty 
>useful. If you like, just add another group of elements to OWL, 
>statements, that are mutually disjoint from classes, instances and 
>relations in OWL DL, but there's a sublanguage that enables us to speak 
>about statements. Or else OWL will suck a lot in comparision to RDF(S) and 
>RDF(S) + Rules will win, because you can't do a lot of the stuff you need 
>to do, like saying what the source of a certain statement is, how reliable 
>this source is, etc. Trust anyone? This is also needed to extend 
>ontologies toward probabilistic, fuzzy or confidence-carrying models.
>
>4) I would love to be able to define syntactic sugar, like partitionOf (I 
>think, this is from Asuns Book on Ontology Engineering). ((A, B, C) 
>partitionOf D) means that every D is either an A or a B or a C, that every 
>A, B or C is a D, and that A, B and C are mutually disjunct. So you can 
>say this already, but it needs a lot of footwork. It would be nice to be 
>able to define such shotcuts that lever upon the  semantics of existing 
>constructors.
>
>5) That said, another form of syntactic sugar - beacause again you can use 
>existing OWL constructs to reach the same goal, but it is very strenuous 
>to do so - would be to define UNA locally. Like either to say "all 
>individuals in this ontology are mutually different" or "all individuals 
>with this namespace are mutually different". I think, due to XML 
>constraints the first one would be the weapon of choice.
>
>6) I would like to be able to have more ontologies in the same file. So 
>you can use ontologies to group a number of axioms, and you also could use 
>the name of this group to refer to it. Oh well, using the name of an 
>ontology as an individual, what does this mean? Does it imply any further 
>semantics? I would like to see this clarified. Is this like named graphs?
>
>7) The DOM has quite nicely partitioned itself in levels and modules. Why 
>not OWL itself? So you could have like a level 2 ontology of mereological 
>questions, and such stuff, all with well defined semantics, for the 
>generic questions. I am not sure there are too many generic questions, but 
>taxonomy is (already covered), mereology would be, and spatiotemporal and 
>dynamic issues would be as well. Mind you, not everyone must use them, but 
>many will need them. It would be fine to find stan dard answers to such 
>generic questions.
>
>8) Procedural attachements would be a nice thing. Like have a a 
>standardized possibilities to add pieces of code and have them executed by 
>an appropriate execution environment on certain events or requests by the 
>reasoner. Yes, I am totally aware of the implications on reasoning and 
>decidability, but hey, you asked what people need, and did not ask for 
>theoretical issues. Those you understand better.
>
>9) There are some ideas of others (which doesn't mean that the rest is 
>necessarily original mine) I would like to see integrated, like a 
>well-defined epistemic operator or streamlining the concrete domains to be 
>more consisten with abstract domains, or to define domain and range 
>_constraints_ on relations, and much more. Much of this stuff could be 
>added optional in the sense of point 7.
>
>10) And not to forget that we have to integrate with rules later, and to 
>finally have an OWL DL query language. One goal is to make it clear what 
>OWL offers over simply adding rules atop of RDF and ignoring the ontology 
>layer completely.
>
>So, you see, this is quite a list, and it sure is not complete. Even if 
>only two or three points were finally picked up I would be very happy :)
>
>Greetings to all from the ESWC,
>denny
>
>
>
>
>----
>Denny Vrandecic
>Institute AIFB, University of Karlsruhe (TH)
>phone: +49 (0) 721 608 6558
>http://www.aifb.uni-karlsruhe.de/WBS/
>blog: http://semanitc.nodix.net

Received on Tuesday, 31 May 2005 16:34:51 UTC