- From: Jeff Z. Pan <jpan@csd.abdn.ac.uk>
- Date: Sun, 8 Jan 2006 20:05:20 -0000
- To: "Peter Patel-Schneider" <pfps@research.bell-labs.com>, "Alan Rector" <Alan.Rector@manchester.ac.uk>
- Cc: <owl@lists.mindswap.org>, <semantic-web@w3c.org>
Hi Peter, Alan and all, After reading Alan's following email and the proposed OWL 1.1 syntax [1], it seems to me that punning is not a convincing choice for metamodeling in OWL 1.1. ( For those who are not familiar with punning - punning means that a name, like Person, can be used as both an individual and a class and a property.) 1. It is impossible to distinguish higher order statements from annotations of symbols and artefacts we are using to represent that domain, as pointed out in Alan's email. The reason that they are not distinguishable is because annotations in [1] are simply syntactic sugar of individual axioms. 2. Datatype axioms, unlike other axioms in OWL 1.1 [1], cannot have annotations. This seems pretty strange, at least to me. The reason is that although individuals, object properties and classes can share names, classes and datatypes cannot. 3.The semantics of punning is not quite intuitive. This can be shown in the following example. In the following OWL 1.1 [1] ontology, Cat and Kitty are used as both classes and individuals. Although Cat and Kitty are the same individual and Ted is a Cat, the ontology Class (Cat partial) Class (Kitty partial) SameIndividual (Cat Kitty) Individual (Ted Cat) does not entail that Ted is also a Kitty. This distinguishes the punning semantics from many other semantics, such as the OWL FA semantics [2], Hilog semantics [3] and RDF semantics [4]. Summary: The above point 2 suggests that the use of punning cannot really eliminate the need for annotations properties. Point 1 provides a good reason to distinguish annotation properties from metamodeling. Point 3 suggests that punning (as an option for metamodeling in OWL 1.1) is not quite intuitive and can be misleading. Greetings, Jeff [1] http://www-db.research.bell-labs.com/user/pfps/owl/syntax.html [2] http://dl-web.man.ac.uk/~panz/Zhilin/pubc.php?type=epapers&id=PHS05 [3] http://dip.semanticweb.org/documents/Boris-Motik-On-the-Properties-of-Metamodeling-in-OWL.pdf [4] http://www.w3.org/TR/rdf-mt/ -- Dr. Jeff Z. Pan (http://www.csd.abdn.ac.uk/~jpan/) Department of Computing Science, The University of Aberdeen ----- Original Message ----- From: "Alan Rector" <Alan.Rector@manchester.ac.uk> To: "Denny Vrandecic" <dvr@aifb.uni-karlsruhe.de>; "Peter Patel-Schneider" <pfps@research.bell-labs.com> Cc: <owl@lists.mindswap.org> Sent: Thursday, January 05, 2006 12:17 AM Subject: Re: [OWL] annotations and meta-modelling in OWL 1.1 > Peter, Danny, All > > Is part of the problem that we have no way to distinguish higher > order statements about the domain from annotations of the symbols and > artefacts we are using to represent that domain? Or do we? > > This seems to me a common problem. When I make an annotation on some > bit of an symbolic artefact, I am giving information about the > symbols in the artefact, not the domain they represents. So if I > author a symbol "Galapagos Finch" and a symbol "Darwin" and a symbol > "Rector", I want to be able to say that the Symbols "Galapagos Finch" > and "Darwin" in this ontology were created by "Rector" but I don't > want to reason about them in the same space. (By contrast if I want > to say that the class denoted by the symbol "Galapagos Finch" was > described by "Darwin", I am talking about the class "Galapagos Finch" > which is an instance of "owl:Class". Or as a more clear cut example, > I want to say that "Galapagos Finch" is an endagered species and be > able to ask queries or form the class of "All members of endangered > species". Again, I am making a higher order statement about the class > "Galapagos Finch" in the domain rather than about the symbolic > representation of that that class. ) > > For annotation of the symbolic representation as opposed to higher > order statements about the class represented, what you seem to need > is a) a means of referring to the symbolic expressions "Galapagos > Finch", "Darwin" etc. and the expressions using them rather than to > their referents. > > Unfortunately, this is not - at least as I understand it - what RDF > Reification does. As I read the Semantics document - please correct > me if I am wrong - the semantics of reification as described by: > > ".. <begin quote http://www.w3.org/TR/rdf-syntax-grammar.>.. > > <ex:a> <ex:b> <ex:c> . > > and > > _:xxx rdf:type rdf:Statement . > _:xxx rdf:subject <ex:a> . > _:xxx rdf:predicate <ex:b> . > _:xxx rdf:object <ex:c> . > > .... end quote" > > is to create a blank node that represents ANY triple of the form > > <ex:a> <ex:b> <ex:c> > > rather than the specific triple in my representation artefact. > > I want to say things about the specific symbolic triple in a specific > resource. I don't know if this is possible by referring to the full > URI of the rdf statement or not, but I can't see how as rdf triples > do not appear to have URIs, even though their subject, property, and > object all do. I think what is being suggested is that OWL axioms, > or at least some OWL axioms, should. > > Obviously, without some kind of partitioning, allowing a statement to > refer to another statement would allow a statement to refer to > itself, directly or indirectly, and would bring with it the > paradoxes of self reference and therefore undecidability, at least > as soon it was used with an expressive enough semantics to include > negation or falsehood. > > The simplest possible form of partitioning would seem to be to > declare that some properties are annotations and do not affect > inference - what I understood annotations did. In fact, what I want > to do, almost always, is query the actual representation with > something like database (datalog) semantics, e.g. "Find all the > axioms in ontology X that were imported from ontology A"; "Find all > the statements in ontologies X,Y and Z asserted to be authored by > 'Rector'" . "Find statements in ontology X for which there is no > assertion of authorship", etc. > > This seems to follow naturally from Ian Horrocks' note on Databases > and Ontologies in which he says "DL KB doesn’t define a single model, > it is a set of constraints that define a set of possible models...In > contrast, DBs (and frame/rule KR systems) make assumptions such that > DB/KB defines a single model" This is a case where I am clearly > interested in one possible model - the symbols and axioms actually in > my implementation of the OWL KB and how they got there, how they fit > together, etc. > > There are a host of issues about copies and versions etc., but I > can't see trying to solve them until the fundamental principles are > clear. To start with at a minimum, I need to be able to send an > artefact along with the description of that artefact in terms of > provenance, usage, and editing annotations needed for somebody else > (who shares the same annotation vocabulary) to use it in the same > way, know where it came from, decide if they trust it, etc. (Note I > trust artefacts not the things the artefacts represent.). > > As far as I can see, The use of the same symbol as a class and > instance in OWL Full is a way of doing higher order representation, > although this is not explicitly clear in anything that I can find. > In which case the problems of fine grained annotation apply equally > to OWL full and OWL-DL. > > In summary, my contention is that there are three distinct kinds of > statements about a) all instances of a class, b) the class itself, > and c) the symbolic representation of the class. Further that we need > to be able to refer to all constructs of the representation that can > be expressed individually including individual triples, axioms, (and > the use of axioms to apply a restriction to a class) as well as > classes and individuals. My concern is that b) and c) are often > conflated. I would refer to b) as "higher order representation" and > c) as" annotation". This conflation is reflected in the common use of > the word "metadata" indiscrimately to refer to either or both. > > I find it hard to see how we can resolve the issues around annotation > definitively without a clear distinction between these two notions. > If this is true, then there is a long term problem for which a best a > 'band aid' can be provided in OWL 1.1. > > Or am I just completely confused? > > Regards > > Alan > > > > > On 15 Dec 2005, at 13:37, Denny Vrandecic wrote: > >> Hi Bijan, >> >> thanks, very good and important questions. I'll try to answer them, >> but I am happy if others would join in resolving them as well. >> >> First, let's ask a question in return: are axioms resources? Do we >> want to be *able* to speak about axioms? Is there anything >> interesting to say about axioms? If you say yes to any of these, >> they need an URI. It appears so totally wrong to me, that we >> (supposedly) can talk about anything and that we ask anyone to >> provide URIs for their stuff, so we can pull it into the Semantic >> Web - but we do not offer URIs for our own statements. I'm against >> a "You shall not annotate axioms"-law. Does anyone disagree on >> that? - and if so, why? >> >>> When is the URI being treated as an axiom? >> >> URIs of individuals, classes, properties... and axioms are all >> disjoint from each other, and thus, just as URIs of individuals, >> classes, properties... etc. axiom URIs need to be either declared >> or it must be inferrable from the context that this is an axiom URI. >> >>> Does the mere presence of an axiom URI in a KB put the axiom >>> itself there? >> >> I don't know. This would mean that the axiom must be known to the >> reasoner. And if it is known, isn't it already there? Hmm. Maybe >> not. I'd prefer a "no, it doesn't", but I am not sure about the >> implications of that. >> >>> Can I make any interesting constructions using the Axiom uri? >> >> I sure hope so, but what do you have in mind? >> >>> Should axiom equivalence entail URI equivalence? >> >> Yes. Because then I can merge ontologies and the annotations of >> axioms merge fittingly. >> >>> (Presumably not, eh, but what about through trivial rewritings, >>> i.e., different rdf serializations?) >> >> Huh, why not? >> >>> Are all axioms named by default, or must one coin a name to talk >>> about them? >> >> The latter. >> >>> Are annotations on axioms restricted to axioms with names/URIs? >> >> Can you make annotations on unnamed descriptions in OWL DL? >> >>> How do we avoid the pain that RDF Reification brings (and it's not >>> just the rdf:subject/predicate, etc. the Statement triple itself >>> and the introduction of the URI means CONSIDERABLE pain....note >>> the fact that no store, to my knowledge, that introduces >>> "contexts" does so by naming all the triples). >> >> I wish I had a convincing answer for that. What about "only >> annotated axioms actually need an URI"? Or "this is just enabling >> you to give an axiom an URI, it is not forcing you"? Or "hmm, I >> think we can hide the pain most of the time, by nicely extending >> the abstract syntax or the OWL XML presentation syntax. It will >> only remain really painful and ugly in the XML/RDF serialization"? >> >> As said, none of these answers really convince me, but maybe they >> are a starter for a solution. Or someone else has one. Maybe there >> is a way to give URIs to axioms without RDF reification, this would >> be great, and I'm just not seeing it. >> >> The question is, is avoiding this pain worth the restriction of >> disallowing to annotate axioms? >> >> Cheers, >> denny >> _______________________________________________ >> OWL mailing list >> OWL@lists.mindswap.org >> http://lists.mindswap.org/mailman/listinfo/owl > > ----------------------- > Alan Rector > Professor of Medical Informatics > Department of Computer Science > University of Manchester > Manchester M13 9PL, UK > TEL +44 (0) 161 275 6188/6149 > FAX +44 (0) 161 275 6204 > www.cs.man.ac.uk/mig > www.clinical-esciences.org > www.co-ode.org > > _______________________________________________ > OWL mailing list > OWL@lists.mindswap.org > http://lists.mindswap.org/mailman/listinfo/owl > >
Received on Sunday, 8 January 2006 20:13:27 UTC