- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Mon, 29 Sep 2003 10:40:34 +0300
- To: ext Bill de hÓra <dehora@eircom.net>, Benja Fallenstein <b.fallenstein@gmx.de>
- Cc: Garret Wilson <garret@globalmentor.com>, Sandro Hawke <sandro@w3.org>, Graham Klyne <gk@ninebynine.org>, <www-rdf-interest@w3.org>
On 2003-09-28 14:49, "ext Bill de hÓra" <dehora@eircom.net> wrote: > > Benja Fallenstein wrote: > > ... >> But I may be wrong-- I think it would be good if some other members of >> the list, in particular members of the core WG, could speak up on how >> literals are to be interpreted in cases like this. > > Any interpretation of a Literal is (must be) determined by the > property it hangs off. Literals do not have a denotation (that's why > they're literal) - although typed literals confuse the matter slightly. I think it's fair to say that any property can can (though need not) affect the interpretation of an object in various ways, regardless of whether the object resource is denoted by URIref, literal, typed literal, or blank node. The property can have rdfs:range assertions which allow one to infer rdf:type assertions about the object resource (which may or may not conflict with explicit type assertions that may exist). The property can have semantics relating to unit of measure which augment a typed literal value, such that the total meaning of a statement may express "ten miles" even if the object alone only denotes the integer value 'ten'. The property may even have extra-RDF semantics known to an application, but not expressed in a way meaningful or accessible to RDF or OWL processors, which results in different interpretations of the same statement by RDF-only applications versus those applications having that extra knowledge about what the property "means". E.g. properties which take plain literal values in the RDF graph, which are actually interpreted as lexical forms of an implicit datatype not referred to in the RDF, such that RDF-only applications will make entailments based on string values that differ from the proprietary applications making entailments based on some typed value. The golden rule, IMO, insofar as portable interchange of knowledge is concerned, is to say as much as you can in RDF, so that little or no special extra-RDF knowlege is brought to bear by any applications on its interpretation. If you don't *mean* the value to be a string, but to actually be something else (other than just the name of that something else) don't use a plain literal. >> I would take the view that a plain literal always denotes a (character >> string, language tag) pair and that properties that can take plain >> literals need to be defined in a way that makes this a reasonable >> interpretation. Typed literals can be used to refer to things like >> integers. > > Exactly. Yup. Cheers, Patrick -- Patrick Stickler Nokia, Finland patrick.stickler@nokia.com
Received on Monday, 29 September 2003 03:40:43 UTC