On 17 Feb 2005, at 20:03, Pete Johnston wrote:
> Yes, but that's one of the fundamental principles of resource
> description on the Web: anyone can say anything about anything. I can
Pete,
Thanks for your comments, I've trimmed all the common ground. I agree
this is one of the fundamentals, and I think I should explain my
motivation more clearly. As you say, the question is about provenance
and my point borders on the usability of derived statements.
I am working on a search engine that indexes DC metadata in Web pages
and stores them as RDF statements. I am leaning to the view that the
direction of the graph has sufficiently different implications for
queries and search results that rev links need to be treated
differently.
My starting position is that users are familiar with the nature of
standard Web search engines, which give results based on the content
of their "target" documents. People soon become aware that publishers
can make inflated statements about the contents of their sites and
make judgements on that basis.
In this case, the provenance of the statements is clear; the subject
URI of the metadata/content is the origin of the statement and
whatever it says reflects the intention of the publisher. To me, this
is equivalent to "rel" statements in the DC schema.
The issue I am grappling with is that people are not used to the
reversal of subject/object relations in search results. I assume that
(as normal) the target of a search will be documents that are the
origin of relevant metadata statements. This is what leads me to wrap
"rev" statements in a statement about their origin.
It would be nice and simple to issue search results for reversed
statements as if they originated from the subject URI, but this would
open up a bizarre new realm of search engine spam!
> There are various approaches to the trust/provenance/context issues
> and reification is one of them (but others here are better placed to
> comment on that work than me!)
I would welcome any suggestions or pointers on this. The indexing
spider will be trawling pages without any fundamental basis for
trust, etc., and only the source URI to work from.
Best regards,
Phil
--
<URL:http://www.codestyle.org/>
|