- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Thu, 25 Mar 2004 18:39:43 +0200
- To: RDF Data Access Working Group <public-rdf-dawg@w3.org>
I've been pondering about the extraction of concise bounded descriptions in the case of "bnode-centric" models such as FOAF, and believe that OWL can provide a very elegant and simple solution to the "overgeneration" of levels that will occur with the present RDF-only definition of concise bounded descriptions. Essentially, owl:InverseFunctionalProperty is employed to bound the depth of descriptions, such that, for any bnode object, if there exist any statements for which it is the subject and which employ properties of rdf:type owl:InverseFunctionalProperty then those and only those statements are included in the description and the recursion ends there. This provides just enough information about the bnode denoted resource for an agent to obtain a complete description of that uniquely identified resource if necessary, in a system-independent manner (irrespective of system-specific bnode labels). In fact, the "fragment description" of the resource consisting of only its uniquely identifying properties constitute a query template, which should result in a complete description of that resource. If no statement with an instance of owl:InverseFunctionalProperty exists for the bnode, then recursion continues, adding all statements about the bnode delimited resource to the description. Now, for vocabularies for which there do not (yet) exist any OWL ontology which specifies which properties are of rdf:type owl:InverseFunctionalProperty, the ability to submit auxiliary knowledge along with an RDF query really shines. If one knows they are querying a knowledge base using a particular bnode-centric vocabulary, they can include along with the query extra information such as, e.g. ex:emailAddress rdf:type owl:InverseFunctionalProperty . or blargh:employeeNumber rdf:type owl:InverseFunctionalProperty . etc. to be taken into account when executing the query, and thus bounding the bnode closure in an effective manner. In any case, one would *have* to continue recursion for bnode denoted resources where no uniquely identifying properties are evident, since otherwise, one is unnable to request a description of that bnode denoted resource in a system-independent, portable manner since bnode labels are system-dependent. So simply limiting the number of levels is not an optimal solution. This OWL augmented approach to generating concise bounded descriptions at least helps to lessen the magnitude of the problem substantially (and completely solves the problem for FOAF, which already defines which properties are instances of owl:InverseFunctionalProperty). Eh? Patrick -- Patrick Stickler Nokia, Finland patrick.stickler@nokia.com
Received on Friday, 26 March 2004 03:20:36 UTC