- From: Elisa F. Kendall <ekendall@sandsoft.com>
- Date: Mon, 16 Jan 2006 09:01:25 -0800
- To: SWBPD list <public-swbp-wg@w3.org>
Hi Gary, Thank you so much for taking the time to read carefully - we really appreciate the feedback. All of your comments are really very helpful, and point out places where we need clarification at a minimum, if not changes to the model as well. I'll forward this to the ODM team independently, and will get back to you with the next version. FYI, and also for Bijan & Peter, since we have asked for their assistance, we have elected to push the final submission out a few weeks so that we can incorporate more changes such as your suggestions and also to allow us to test the QVT mappings. We did not want to publish a document containing "mapping code" that had not been tested with the parsers at a minimum, and only recently did such a tool become available. Thanks again, Elisa Gary Ng wrote: >Elisa, > >Sorry for the delay in sending these out. I took some extra time to read >it more carefully. > >I have mainly focused on the OWL Chapter of [1], with some look ups into >the DL and RDF Metamodel chapters. The original reference to the doc is >at [2]. > >The OWL descriptions are accurate. I believe it has covered all OWL >constructs with minor quibbles. See "Quibbles" below with minor change >suggestions. > >Recap: Base on the OWL abstract syntax, the metamodel covers all >language elements of OWL (OWLBase). The base model is a minimally >constraining specification. The sub-language variants are achieved per >application/vendor by constraining the usage of the elements, provided >as additional contraint packages. The OWLDL and OWLFull Constraints >Package have yet to be defined and I understand Peter Patel-S and Bijan >are helping to define those. > >Comment: Generally I find the interaction between RDF Document, OWL >Graph, RDF and OWL statements, together with OWL Universe, needs some >clarification. > >My concern can be summarized in the following questions, I find that the >answers are probably implied within the text, but I haven't found >satisfactory answers by myself and I need some help: > > >1. It seems OWL Universe, OWL Statements and OWL Graph are meant to be >derived objects that tool vendors are to support? No end-users will need >to explicitly specify them? > > >2. In 12.2.2: "As shown in Figure 14, an OWL ontology consists of a >collection of facts, axioms, and annotations, defined in terms of zero >or more RDF graphs and statements." > > Can statements made in RDF that represent OWL Expressions readily >translatable into OWLBase constructs? Can a resource graph specified in >the RDF profile be readily usable as an OWL instance graph? > > From the other direction, RDFGraph contains RDF Statements. >OWLGraph is a subset of RDFGraph. Not all graphs are valid OWL Graph. >But does this mean that all OWL Expressions are also RDF Statements, >just like the relationship between RDF and OWL syntax? Need to clarify >that whether axioms and expressions made using OWL Metamodel has an >equivalent RDF Statement form just like the language. > > But if the OWL Metamodel is based on the abstract syntax, which >means it does not necessary correspond to the RDF syntax for axioms, >which in turn would mean the OWL Ontology contains RDF Graphs + some >objects which can only be accessed through OWL Universe? > > >3. There seems to be two ways to specify an RDF triple? > > One is to use a Statement stereotype, the other is to literally >associate one object to another with a link. The former allows >reification, and associate the statement with a graph, the latter >doesn't, and the fact that it becomes a statement is actually implicit. >Or is it the case that all triples must be specified through a statement >object with 3 links to subject, predicate, object explicitly, such that >reification and or named graphs can be specified? > > >4. Not really about the OWL MetaModel, but will we expect a rewrite of >the DL MetaModel (Appendix C)? What is expected to be done with it? It >looks like it still needs a lot of work. > > >============================ Quibbles ============================ > >Naming consistency: Suggest the names RDFGraph (for RDF) and >OWLStatement (for OWL) rather than Graph (for RDF) and Statement (for >OWL). Thus it is clear that there are: RDFGraph, RDFStatement, OWLGraph, >OWLStatement. > >12.3.6 Individuals > >SameIndividual and DifferentIndividual with associations sameIndividual >and differentIndividual have not been introduced before or anywhere else >in this chapter. > >12.3.11 OWLDataRange > >OWL DataRange can also be the rdfs:range of a Property but is missing >here. It is certainly not in the RDF Metamodel and thus should be >included here. > > >12.3.13 RDFProperty > >rdfs:domain and rdfs:range of property is covered in the RDFMetamodel >and is not covered here. In the spirit of completeness, perhaps a >pointer here to section 11.5.1 where they are described would be good. >Since they are also an integral part of OWL. > > >12.4 OWLBase Property > >Under the meta class Property, the siblings specified here are: >FunctionalProperty, DatatypeProperty, ObjectProperty; Under >ObjectProperty, there are InverseFunctional, SymmetricProperty, and >TransitiveProperty. > >I thought in OWL-Full DatatypeProperty may be InverseFunctional? The >organization certainly suggests that if it is inverse functional, it is >automatically object property, which is the case in DL. I went to >checked the webont reference: > > "NOTE: Because in OWL Full datatype properties are a subclass of >object properties, an inverse-functional property can be defined for >datatype properties. In OWL DL object properties and datatype properties >are disjoint, so an inverse-functional property cannot be defined for >datatype properties. See also Sec. 8.1 and Sec. 8.2." > >(thinking aloud here...) So to create an inverse functional Datatype >property for OWL Full ontology from this meta model, one will have to >declare that property as both inverse function and as datatype. As >inverse functional prop is in turn an object property... the prop is >both a datatype and object property => which is consistent with RDF and >OWL-Full semantics. > >Granted. No issue. But perhaps to include some explanation to reassure >the reader that the semantics is correct with reference to WebOnt >reference? > > >12.6 Datatypes > >Perhaps to explicitly mention that the OWL's notion of unsupported type >is automatically handled too? > >Doesn't need much description, as TypedLiteral and the RDFDatatype are >already described in the RDFMetaModel. Again, this is more like just >covering all OWL constructs like a checklist in this chapter rather than >just leaving it implied in the TypedLiteral section. > > >12.2.3 RDFSLiteral > >This section seems out of place and interrupted the flow. Perhaps move >to before or after 12.3.15 TypedLiteral? > > >12.3 Class descriptions > >The alphabetical ordering of the sections on various annonymous classes: >"ComplementClass, Enumerated, Intersection, Restriction, Union", could >perhaps be arranged from simple (complement, intersection, union) to >complex ideas (restriction)? It was slightly interrupting reading >Intersection, then a detail long Restriction section, then back to >Union, which carries similar content to Intersection. > > >Typos >----- > >12.1 > > "Vendors who are interested in supporting OWL Lite would simply use >only the relevant constructs from the base package and tighted a few >constraints from the OWL DL package, as required." > >Change "tighted" to "tighten" perhaps? > > >[1] http://www.omg.org/cgi-bin/doc?ad/05-09-08.pdf >[2] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/0174 > > > >
Received on Monday, 16 January 2006 17:01:38 UTC