The first layer of SemanticWebArchitecture consists of basic facts like "The sky is blue" written for machine consumption as logical formulas. For noun phrases ("terms" in logic-speak) we can use
- URIs somewhat like proper names,
- BlankNodes somewhat like the pronoun "something"
- strings and other literal expressions
The WebOnt and RDFCore WGs are all snarled up over the details of using XML language tags in text values, and the details of integrating text values with datatype expressions. Discuss. ;-)
seems somebody independently started RdfCoreXmlLiteral. PPR:RefactorMe.
For exampleFootNote(using ShorthandRDF):
<theSky> <color> _:something. _:something <colorName> "blue".
String literals can be qualified by language tag. But what does that mean, formally? rdfms-xmllang/test007a.nt from RDF test cases shows that a string tagged as french is distinguishable from a string tagged as english; i.e. it's not specified to denote the same thing:
<http://example.org/node> <http://example.org/property> "chat"@fr .
does not entail
<http://example.org/node> <http://example.org/property> "chat"@en .
The relationship between plain literals and strings of type XML Schema string is interesting...
_:x rdf:value "abc".
_:x rdf:value "abc"^^xsd:string.
catalog:productDescription rdf:type owl:FunctionalProperty; rdfs:comment """it's important that various copies and parts of this sort of product catalog describe products consistently""". <#item1> catalog:productDescription "<p>nifty doodad</p>"^^rdfs:XMLLiteral. <#item1> catalog:productDescription "<p >nifty doodad</p>"^^rdfs:XMLLiteral.
is that an inconsistency? Note that RDF entailment and OWL consistency are parmeterized by a set of datatypes; with the rdfs:XMLLiteral datatype in the set, this is consistent; the objects of both triples denote the same value. hmmm... it's consistent, i.e. satisfyable, even without that datatype. Hmm... need a better test case after all.
We can also use datatype expressions (a limited form of FunctionalTerm), but those are built from URIs and text values.
Fodder, newest first:
- 20 Mar 2003: WebOnt telcon record, including datatype decision and my action to get review of it
- Feb 28 2003 just strings, please
- RDF core issues: charmod-literal, etc.
- Jan 2003 ISSUE: XMLLiteral and xml:lang raised in WebOnt
- May 2002: 5.8 Datatypes WebOnt issue raised: XMLLiteral, etc.
- IESG note on human text vs. protocol strings (LinkMe)
- RDF '99 spec, esp section 7.5. Values Containing Markup
hmm... perhaps xml:lang is just not the right tool at all... consider TopicMaps scopes and such... PropertiesForNaming