- From: Jim Hendler <hendler@cs.umd.edu>
- Date: Fri, 7 Feb 2003 09:59:34 -0500
- To: webont <www-webont-wg@w3.org>
I believe that annotations are quite important, in fact mandated by our charter >Within the communities of researchers working on the creation of the >semantic web, there are many different visions of what ontologies >are and how they can be used. This working group will strive to be >as general as possible in supporting multiple approaches and >techniques. here are some use cases as to why I think annotations are needed for the above 1 - Ontological properties we didn't think of: there are clearly reasonable kinds of semantics that we as a group didn't address. A good example of this would be temporal logics -- one could easily imagine a later group of researchers deciding they'd like to extend OWL into OWL-TIME (as some are doing now for DAML [1]). One can easily imagine a language extension such as: :InitialEvent a owl:class; owl-time:ComesBefore :finalEvent; owl:equivalentClass ... . This would add the extra triple into the RDF graph that these researchers would need to do their thing. I imagine other ontologies would want to do this sort of thing with part-whole or other such - clearly we don't want owl-time:ComesBefore and owl-time:ComesAfter to entail each other. Note also that some people may use annotations like this that might not be fully definable in a logic, but could be crucial to e-business and other web application. I think our language could be considered seriously broken if, for example, we were doing digital rights management (DRM) and our design said :VaccinationRecord a owl:class; drm:permission drm:private. entails :VaccinationRecord a owl:class; drm:permission drm:public. Since we are not omniscient and able to guess all possible of these, I believe we need to allow them as annotations. (Note: Drew McDermott wrote a seminal paper many years back called "Artificial Intelligence meets Natural Stupidity" that talks about confusing labels and meaning -- those on the WG who have never read it, should seek it out) 2 - RDFS labels needed for internationalization, language and other non-logical features: if lexeme:Embarass a owl:class; rdfs:label "(Lang:US/En)". entails lexeme:Embarass a owl:class; rdfs:label "(Lang:IB/Sp)". then every time our faces turn red (root "Embarass" in English) we will discover ourselves pregnant (root "Embarass" in Spanish) similarly "rdfs:Label "(Display in Blue)"" shouldn't entail "rdfs:Label "(Display in Red)"" 3 RDFS comments RDFS comments serve the same purpose sometimes espcially as they are poorly named. If instead of calling them rdfs:comment they had been called rdfs:stringToBeAssociatedWithThisGraph then we would not expect them to be ignored as Peter and Ian were arguing (correctly) that comments generally are. Comments that one wants an rdf parser to ignore are put in documents using XML comments or the like, However the "rdfs:comments" may not be machine understandable but keeping them separate from a entailment point of view can be important. I certainly don't expect the system to recognize what the different meaning of different comments is, but I would expect the fact that the graph includes triples that have different RDF semantics would mean someone did something on purpose to put those in the graph - otherwise they would use XML comments - for example, the rdfs:label cases above could have been done with rdfs:comment. (note: I would not like, but could live with, a solution that suggested in our documents that users should not use rdfs:comment unless they explicitely want to have them effect the meaning, and recommends use of XML comments instead - ugly, but I could live with) -JH [1] http://www.cs.rochester.edu/~ferguson/daml/ -- Professor James Hendler hendler@cs.umd.edu Director, Semantic Web and Agent Technologies 301-405-2696 Maryland Information and Network Dynamics Lab. 301-405-6707 (Fax) Univ of Maryland, College Park, MD 20742 240-731-3822 (Cell) http://www.cs.umd.edu/users/hendler
Received on Friday, 7 February 2003 09:59:38 UTC