- From: Dan Brickley <danbri@w3.org>
- Date: Thu, 5 Aug 2004 07:07:23 -0400
- To: "Miles, AJ (Alistair) " <A.J.Miles@rl.ac.uk>
- Cc: "'public-esw-thes@w3.org'" <public-esw-thes@w3.org>
* Miles, AJ (Alistair) <A.J.Miles@rl.ac.uk> [2004-08-05 11:54+0100] > > Just to add something here on an issue that came up on the wiki ... > > [From the wiki] > A top concept is defined by the absence of skos:broader terms. One can > easily determine that whether a concept is a top concept. > > ... i.e. arguing that no explicit mechanism for identifying top concepts is > required. > > I think there are two counter-arguments to this, the first from Stella's > mail today ... Yes, both those arguments hold. More generally, in RDF we don't have a "negation as failure" approach. The absence of information from some graph doesn't mean it isn't true, just that it isn't said. This is part of the "open world" design behind RDF, which emphasises merging of scattered datasets and partial knowledge. So RDF vocab definitions oughtn't to be couched in terms of the characteristics of _RDF descriptions_ of the things they define. When we define "top concept" in terms of absence of bits of information from some RDF graph (ie. info about broader terms) we get tied in knots since it is not clear _which_ RDF graph. And you can always omit things from some RDF and be truth-preserving, since all the statements in the graph are AND'd. So yeah, it seems "top concept" is motivated. cheers, Dan
Received on Thursday, 5 August 2004 07:07:23 UTC