Re: question about imports

pat hayes wrote:

> The point of this is that one intuitive argument that has been given for
> the use of owl:imports is that the cross-references between ontologies
> might support a kind of semantic Google process whereby the most-imported
> ontologies are rated as more trustworthy, or something like that. But since
> the current semantics doesn't require the links to exist, that seems to
> mean that it fails to capture something of potential importance. Roughly,
> the thing that is missing is a relationship between documents (or document
> tokens, or ontologies, maybe) which is that one of them agrees with, or
> endorses, the other.

I'm not troubled by this at all, for the following reason.

Unless I'm very wrong, semantics only deals with the "propositional content" 
(what is asserted to hold in a set of sentences and what follows from what is 
asserted).

The issue you raise seems more a matter of pragmatics (how to use a set of 
sentences for various non-truth-preserving actions, like assigning trust, 
convincing someone, etc).

Although I'm far from an expert on stuff like this, I thought it was well 
accepted that two sets of sentences which have the same propositional content 
(ie are semantically equivalent) might well carry different pragmatic 
implications.
Your example seems just an example of this:

if doc A makes 10 statements,
and I'm not familiar with the source of doc. A
I have no particular reason to trust doc A.

if I'm not familiar with the source of doc AC,
and doc AC makes 1 statement,
plus it imports the 9 other statements from a source I trust,
I may have more reason to trust AC's content than I do for A's content,
even though they are semantically equivalent.

I'm quite happy for the world to be like this,
(in fact, the real world is like this),
and don't see a problem.

Frank.
   ----

(Another example just occurred to me:
if I find the same content in 100 different places on the web,
from different independent sources,
I'm more likely to believe it then when I find just a single copy,
even though the additional 99 copies are just semantically equivalent to what 
I knew already after the first copy).

Received on Thursday, 5 December 2002 17:07:58 UTC