On Thu, 19 Feb 1998 [log in to unmask] wrote:
> I would probably want the primary content meta data to include "recipe",
> "cheese" and "bread"; but if I include beer, I'm going to pull in an awful
> lot of people who have _no_ interest in my recipe (people looking for beer
> "joints", beer company's, beer recipes, etc.) Only some very small
> percentage of the people who put "beer" in their query are actually looking
> for my recipe.
I think you'll find that many people that currently add metadata to web
pages will include the "beer" keyword. Not to mention "sex", "hot girls",
"Pamela Anderson", "gential" and about a thousand other keywords that bear
little or no relation to the content of the resource. For them "false
positives" in the search engines just mean more hits and more advertising
revenue (and hey, maybe a few nerds that stumble onto their pages will buy
an overpriced subscription to the members area with the low quality
MPEGs). So I don't _think_ it has been suggested to have multiple tiers
of subject descriptors in DC per se. However, you can always use (or
invent if nothing is appropriate) a scheme that has such a tiering and
then use it in your DC metadata. If search engines and tools (such as
yours) understand it they can do something useful with it and if they
don't they probably revert to just seeing it as a flat list of terms.
What is need is some way of signifying the quality, or at least source, of
the metadata. I guess PICS could come in here and you really want strong
digital signatures with non-repudiation (so that the spamsters can't say,
"Ooooh no m'laud, it wasn't me that falsely advertised my pages with
metadata of no relevance to the subject at all. It must have been one of
those teenage hackers"). We might (indeed I'll stick my neck out and say
definately will) even want to have many resources with multiple sets of
metadata with different qualities. Definately sounds like a Warwick
Framework thing to me.
Tatty bye,
Jim'll
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Jon "Jim'll" Knight, Researcher, Sysop and General Dogsbody, Dept. Computer
Studies, Loughborough University of Technology, Leics., ENGLAND. LE11 3TU.
* I've found I now dream in Perl. More worryingly, I enjoy those dreams. *
|