- From: Jim Hendler <hendler@cs.umd.edu>
- Date: Tue, 18 Feb 2003 16:30:59 -0500
- To: Mike Dean <mdean@bbn.com>, Guus Schreiber <schreiber@swi.psy.uva.nl>
- Cc: webont <www-webont-wg@w3.org>
here is my detailed review of the Reference document. As I said previously, the Reference is much improved, and this long list of comments only talks to the editorial issues - technically the document is now quite strong. -JH general typographical note -- some language features appear in mixed upper lower case font others appear in "small capital" font. Editors should pick one or the other and make these consistent. Sect 1.1 Purpose of the document - 1st paragraph copyedit: should end "for a more narrative description and examples of the use of the language. 3rd pp: would be best if you mentioned all documents - you should include a mention of use cases and Test. Section 1.2 (and throughout) -- you are uneven in your use of links to other documents. You should make sure that all mentions of Overview, Guide, etc. are linked. 3rd paragraph (optional) this would be a good place to say something about Owl "flite" -- that is to mention that Owl Lite is a vocabulary useful in two ways - as a subset of the Full Owl and as a subset of DL, and that we will use the term "Owl Lite" for the latter. 1.3 copyedit Thus, it is allowed => Thus, it is allowable final sentence copyedit and is therefore perfectly allowed => and therefore would convey the same meaning. 1.4 - editing suggestion I think this section needs to be longer, and we need to discuss more specifically the upgrade path from RDF(S) to OWL -- we have to address the fact that RDFS with OWL is not usually going to be OWL Lite, otherwise we will get adverse comments (cf. CG discussions at www-semweb-cs@w3.org - member only) 1.5 annotations Now that we have a solution, we need to add it here. This will be the primary place where it is discussed. We also have to make it clear how reference to other ontologies (without imports) works -- that is, do these need to be types as annotations or not? 1.6 change second paragraph to (copyediting): The examples in this document are meant to serve as illustrations of the use of OWL language constructs. They do not form one consistent ontology. For an extended example the reader is referred to the Guide Document [ref]. 2.0 intro -- this needs to be reorganized -- some of it belongs back in section 1. I'd suggest creating a section that describes the appendices of this document, and then put the rest in 2.0 where it belongs. Thus, I suggest the following becomes section 1.7, and then section 2.0 is the rest. (about rdf:RDF etc.) ------------ HTML FOR CHANGES TO SECTION 2.0------------ <h3><a name="AppendixList">1.7 Appendices to this Document</a></h3> <p> This document has a set of appendices containing additional information. Links in the text attached to definitions of language constructs provide access to the OWL Abstract Syntax and Semantics [<cite><a href="#ref-abstract-syntax">OWL AS&S</a></cite>]. <a href="#appA">Appendix A</a> contains a systematic set of links for each language construct to the corresponding sections in the Guide and the AS&S documents.</p> <p><a href="#appB">Appendix B</a> contains an RDF schema for the OWL language constructs. This schema defines the OWL language constructs in terms of RDF Schema classes and properties. This schema provides the basis for the RDF/XML syntax of OWL. Conventionally, classes have a leading uppercase character; properties a leading lowercase character. Thus, <code>owl:Ontology</code> is a class, and <code>owl:imports</code> is a property. </p> <p> <a href="#appC">Appendix C</a> gives a tabular overview of all OWL language constructs in terms of the built-in OWL classes and properties (the latter with their domain and range).</p> <p> The <a href="http://www.w3.org/2001/sw/WebOnt/charter"> Web Ontology Working Group Charter </a> states that <blockquote cite="http://www.w3.org/2001/sw/WebOnt/charter"> The Working Group shall start by evaluating the technical solutions proposed in the DAML+OIL draft. If in this process the Working Group finds solutions that are agreed to be improvements over solutions suggested by DAML+OIL, those improved solutions should be used. </blockquote> Thus, the <a href="http://www.w3.org/TR/daml+oil-reference">DAML+OIL (March 2001) Reference Description</a> provided the basis for OWL. For readers familiar with DAML+OIL, <a href="#appD">Appendix D </a> lists the changes between DAML+OIL and OWL.</p> ------------ End of HTML FOR CHANGES TO SECTION 2.0------------ 2.1 As well as talking about the name space we need to say something about Mime type. We seem to have lost this as an appendix, but not suggested it as a separate document ... something needs to happen in the WG, and be reflected here. 3.0 copyedit owl:Ontology a ontology => an ontology suggestion the ontology header uses the example "owl:imports ..." and imports the owl namespace doc. I suggest we use a different example or do two imports, to make it clearer that any ontology can be imported, not just the name space. owl:imports - reword The text as written is confusing. I suggest instead Note that the importing a document is different than creating a namespace reference. owl:imports do not set up a shorthand notation for names as does a namespace reference. On the other hand, the namespace reference does not imply that all (or even any) ontological terms from that namespace are being imported. Therefore, it is common to have a corresponding namespace declaration for any ontology that is imported. owl:versionInfo - there is an "@@" in here, should be removed last sentence in owl:backwardCompatibleWith - copyedit and thus has also => and thus also has section 4 - classes there is a missing example (the @@) - I'm not sure an actual example is needed here, suggest deleting. However, I think we need some test in here that explains when to use owl:class -- doesn't have to be technical, just something that makes it clear. For example, could be as simple as In OWL Lite and OWL DL, owl:class must be used for all class definitions. In OWL Full, owl:class and rdfs:class are synonyms and either one can be used in a class definition. 4.1 - copyedit ...which satisfy the property restriction (3rd type) => ...which satisfy the particular property restriction (3rd type) 4.1.2 property restrictions, sentence about owl:Restriction -- it says "c.q. cardinality restriction ..." - I think c.f. is better here, but whichever you use, make sure it is in italics to indicate that this isn't a typo. end of section 4.1.2 says "see the section on properties" -- this needs to be a link. 4.1.2.2 cardinality restrictions it reads "The default cardinality of properties is "any". that is unclear, but more importantly this is an important point that needs to be made. How about In OWL, it is assumed that any instance of a class may have an arbitrary number (zero or more) of values for any property. To make a property required (at least one), to allow only a specific number of values for that property, or to insist that a property must not occur, cardinality restrictions must be used. OWL provides three constructs for restricting the cardinality of properties (locally) within a class context. owl:maxCardinality I'd like to suggest not using two parents in both example -- why not make the first example "at least one parent" I'd also suggest we end that section by adding a sentence which reads Note that an owl:minCardinality of one or more means that a value for the property is required of any instance of the class. owl:intersectionOf the WG decision to include owl:intersectionOf in Lite is not reflected in this section - it appears in a note in a later section (on axioms for complete classes). I'd suggest that we introduce a note in this section that simply says NOTE: OWL Lite is restricted in its use of owl:intersectionOf. This is discussed later in this document, see section (appropriate name/link) There is a NOTE about the unique names assumption that occurs in this section (4.1.3 in owl:intersectionOf). I think the discussion of the unique names needs to be moved elsewhere (probably with allDifferentList) and this should just point at that. owl:unionOf and owl:complementOf these are not available in OWL Lite, this should be noted. example of owl:complementOf - since you are reconsidering anyway, I think it would be less confusing to use a simple example of just "not meat" and use a hasValue instead of a collection. Mixing the union and the complement blurs this example anyway rdfs:subclassOf There is an @@ asking whether to add a note about rdf:about v. rdf:id -- I would suggest NOT adding that - no reason for it here. I would suggest removing the discussion of multiple subclass axioms for the same class defining an intersection-- and take out the example of the two semantically equivalent definitions of Opera. This stuff is in Guide, and doesn't seem to be needed here. owl:equivalentClass has an @@ that should be removed The note at the end of the section on owl:equivalentClass - copyedit As OWL should be usable in a distributed environment, this may be a useful feature. => As OWL is usable in a distributed environment, this can be a useful feature. Axioms for complete classes without using equivalentClass the note at the end of this section about OWL Lite mentions the use of intersectionOf. this is the note that should be referred to in the intersectionOf section (see above) the @@ asks whether to add a guideline on explicit v. implicit form of complete classes -- I wouldn't add anything else at this point, so that could be deleted. SECTION 5 - Properties remove the parenthetical remark about a possible better term for object properties -- silly for a reference to say "there's a better name, but we didn't use it" copyedit - characteristics of a properties => characteristics of a property sentence following <owl:ObjectProperty rdf:ID="hasParent" /> copyedit -- say "this defines a property with the restriction (by default) that" (i.e. add the "by default") NOTE at end of rdfs:subproperty says "in OWL DL the range and domain ... should be either both ..." but that should be "must be either both" as this is mandated in the design (optional) I wonder if the section on subproperties might be a good place to mention that owl's use of local property restrictions adds power -- i.e. in rdfs it is easy to say that hasMother is a subproperty of hasParent, but very hard to say that if the class is "Kitten," then hasMother is restricted to be of class "cat" -- might be worth just pointing to local properties here and saying that sometimes local properties can be used to avoid creating artificial subproperty hierarchies (i.e. having to create KittensMother subproperties) rdfs:range there is a note at the end that rdfs:range restrictions should be used with care. I believe the same is true of rdfs:domain restrictions (although maybe less so) - consider a similar note in the domain section (Optional) owl:equivalentProperty I don't see a need for an example in this section, the concept is clear. I'd suggest just dropping the @@ add example. owl:inverseOf - there is an @@ for a note that does, indeed, still need to be added. owlFunctionalProperty there is a sentence which reads "This corresponds to the notion of "optional" value encountered in many data-modeling notations" - maybe I'm dense, but I don't see the relation between functional property and optional value. This should either be deleted or explained better for dummies like me. owl:inverseFunctionalProperty you use the example of social security number, but that's not a really good one as it is a one-one relation -- I'd like to suggest an example where it isn' t one to one. possibiltiies would be the mapping from children to mothers -- i.e. "BiologicalMother" is inversefunctional -- cannot have BiologicalMother(Mary,Jim), Biological Mother(Jane, Jim) - however, My mother can (and does) have several biological children - so this is an example of an inverseFunctional property that is not also functional (which is the problem with SocSec #) owl:SymmetricProperty - copyedit change first sentence to "The OWL property class owl:SymmetricProperty is also a subclass of owl:ObjectProperty. note: is this true? Could we have a property in OWL that is symmetric but has datatypes for range and domain? Mathematically this would be useful (i.e. the class abelian group is a subclass of group for which certain symmetry properties between members, which must be datatypes, hold). If we don't allow it in Lite or DL, do we allow it in Full? 6.0 individuals We should state explicitly that individuals are typically defined using RDF and that the examples we provide are just for additional clarity and to show how OWL can be used to add some additional expressivity to the RDF. Add a sentence to that effect in section 6.0, and then go to 6,1 (individual axioms" and it would be much clearer. 6.2 owl:sameAs and owl:sameIndividualAs it would be worth adding a sentence at the end contrasting these to equivalentClass. i.e. the example of football team being same as soccer team could be contrasted to <:footballTeam owl:equivalentClass us:soccerTeam /> to make clearer the difference between denoting same thing and denoting same extension. owl:AllDifferent I think we need to emphasize that owl:distinctMembers has no meaning outside the context of owl:AllDifferent. (add a NOTE to this effect?) - explanation of how to add to the list is important - should be added (or referred to if Guide adds it) 7.2 enumerated datatypes we need to keep an eye on RDF for this one. I, and possibly others, have raised it in their comments list. If they decide to allow datatypes in collections, then we will have a nicer syntax for this powerful feature. 8.0 the section refers to an "informal" description. Let's make that "informative" which is still true, but less pejorative sounding as this is now the only place in our non-normative documents where we really discuss this stuff, particularly the implication that OWL Lite vocabulary in Full might be a useful subset 8.1 might be a good place to mention that RDFS documents will generally be in OWL Full unless they are specifically constructed to be in DL /Lite 8.3 the discussion of OWL Lite doesn't mention that there is a complexity class reason for it as well. This section also doesn't discuss compatibility with RDFS. I think these are easily fixed: i. add a sentence at the end of the paragraph starting "The idea behind the OWL Lite expressivity limitations..." which states In addition, the limitations on OWL Lite also place it in a lower complexity class than OWL DL. This can have a positive impact on the efficiency of complete reasoners for OWL Lite. ii. drop the first sentence of the last paragraph (the one which starts Owl Lite is constrained to be a subset of Owl DL" because it is redundant. Start the paragraph with the following sentence with the word "Implementations" (i.e. delete "In addition" from the start of that sentence. iii. add RDFS models to the list of things Lite (without DL) could be used for -- i.e. in providing interoperability of OWL systems with RDFS models, databases, ... iv. change the very last sentence to "The Web Ontology Working Group has not provided a name for this potentially useful subset." 9, Class and property deprecation - copyedit delete the words "Once again" from the second to last sentence in the first paragraph of this section Appendix A We don't list the rdf/rdfs constructs used in OWL -- this makes sense because this is a linking table to our own documents -- but maybe add a sentence that just says "for any feature from RDF or RDF Schema please see the appropriate documents" with a link to our references which in turn link to their documnts Appendix D Something probably needs to be said in here about the fact that we have changed the semantics to a large degree, and also that we added the three sublangauges with D+O being most like Owl DL. -- 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 Tuesday, 18 February 2003 16:31:18 UTC