- From: <herman.ter.horst@philips.com>
- Date: Fri, 28 Nov 2003 15:43:28 +0100
- To: horrocks@cs.man.ac.uk
- Cc: www-webont-wg@w3.org
As was discussed in the last telecon, here is an overview of changes in S&AS required in view of changes in the RDF Semantics document. I only consider the main text, Sections 1 to 5. Many of the changes I describe below propagate into the proof appendix. === Section 5.2, second paragraph. The changes needed in this paragraph can be summarized as follows: -delete parenthetical sentences referring to RDF Core's last call -a vocabulary also includes literals, even plain literals -not all plain literals are needed in LV (the assumption that all plain literals are in LV led to horrible errors and complications in the RDFS Herbrand construction and the proof of the RDFS entailment lemma; analogously this generalization will simplify formal considerations with OWL interpretations) -instead of a "datatype theory T", the RDF Semantics speaks of a "datatype map D". This requires many textual changes in S&AS, which also solve the confusion in S&AS with the mapping to RDF graphs, which is also called T. -LI maps to RI and not to LVI. This is needed, for example, for ill-typed XML literals, and also for other ill-typed literals. -The current text of S&AS gives RDF Core's last call definition of the function CEXTI, which has been made more explicit. Instead of describing the needed corrections more precisely, it takes less effort to suggest possible new text for this paragraph, keeping the current words as they are as much as possible: "From the RDF Semantics [RDF MT], for V a set of URI references and literals containing the RDF and RDFS vocabulary and D a datatype map, a D-interpretation of V is a tuple I=<RI,PI,EXTI,SI,LI,LVI>. RI is the domain of discourse or universe, i.e., a nonempty set that contains the denotations of URI references and literals in V. PI is a subset of RI, the properties of I. EXTI is used to give meaning to properties, and is a mapping from PI to P(RI x RI). SI is a mapping that takes each URI reference in V to its denotation in RI. LI is a mapping that takes each typed literal in V to its denotation in RI. LV is a subset of IR that contains at least all plain literals in V. The set of classes CI is defined as CI = { x in RI | <x,SI(rdfs:Class)> in EXTI(SI(rdf:type)) } and the mapping CEXTI from CI to P(RI) is defined by CEXTI(c) = {x in RI | <x,c> in EXTI(SI(rdf:type))} for c in CI. D-interpretations must meet several other conditions, as detailed in the RDF semantics. For example, EXTI(SI(rdfs:subClassOf)) must be a transitive relation and the class extension of all datatypes must be subsets of LVI. " Other changes, in the next definition: - replace "let T be a datatype theory..." by "let D be a datatype map..." - in the same definition, the 5-tuple should be made a 6-tuple that also includes LVI - in the same definition, D-interpretation instead of T-interpretation - In Section 5.2, it should be said that an OWL interpretation, and also an OWL Full interpretation and an OWL DL interpretation, are defined with respect to a datatype map D. The same is true for OWL Full entailment and OWL DL entailment. This is also made explicit in this way for abstract OWL interpretations and abstract OWL entailment in Section 3. Also, the statement of the correspondence theorem needs to be made more explicit in this way. === Section 3.1 Several corrections in order to make the details of this section parallel/consistent with those in Section 5. Below I suggest additions in the text with /.../: -first sentence: "...the vocabulary must include all the URI references /and literals/ in that ontology ... but can include other URI references /and literals/ as well." -Definition: "An OWL vocabulary V consists of /a set of literals L/ and seven sets of URI references: ..." (It seems that the letter L is still free.) -again: replace datatype theory T with datatype map D (the definition given here matches exactly with the RDF Semantics document). -Definition: Let D be a datatype map. An abstract OWL interpreation w.r.t. T with vocabulary L, VC, etc. -It is now assumed that LV contains each Unicode string and each pair of two Unicode strings. For the correspondence with Section 5, it would be sufficient to assume that only plain literals in V (and L) are contained in LV. -"L:TL->LV, where TL is the set of typed literals /in L/" (As a side-effect advantage of this correction, the set TL doesn't fall out of the air anymore.) -"S is extended to plain literals /in L/ by (essentially) mapping them onto themselves..." -In order to make the definition more strongly parallel to the definition of ill-typed literals (and datatype clash) in the RDF Semantics document, it seems to be better to replace, in the last bulleted condition, R-V(d) by R-LV Other remark about Section 3.1: There are now two definitions of the function S, one with domain VI and the other with domain VC union etc. It is confusing that these domains are not explicitly assumed to be disjoint, although I can see what the intention behind this recently made correction is. It seems better to handle this intention as follows: Define S just once, with domain VI union VC union ... and add the assumption S(VI) subsetof O. === Section 3.4 Definition: -datatype map D instead of datatype theory T. -add L === Section 2.2, abstract syntax: -The text uses the correct word "plain literal", but the productions speak (just twice) of "untypedLiteral", which should be replaced by "plainLiteral" -Section 2.1 speaks of "other built-in RDF Schema datatypes which are problematic for OWL", which should be XML Schema datatypes. -Section 2.1 speaks of Section 3.4 of RDF Semantics, which does not exist. It should be Section 5.1. === Section 4: -there are two references to "RDF/XML graphs", which should be "RDF graphs". (The references to RDF/XML documents are correct, of course.) === Section 2.3 "as has been made evident in several discussions in the WG". Shouldn't this very context-dependent reference, to WebOnt, be removed from the document? === -typo: individiual-valued -Section 2.1: sentence "The class with ... individuals" does not end with a period. === As I said, many of the changes I describe here lead to the need for corresponding changes in the proof appendix. === When I compiled this list, I noted the need for a small correction to the RDF Semantics document, which I sent to rdf-comments [1]. Herman ter Horst [1] http://lists.w3.org/Archives/Public/www-rdf-comments/2003OctDec/0212.html
Received on Friday, 28 November 2003 09:44:09 UTC