Lists Home |
Date Index |
Didier PH Martin wrote:
> <rdf:description about=http://mydomain.com/myresource">
> <category>weight loss</category>
> Considering the huge market for this category (especially in the USA) do you
> firmly believe that this meta-data will trustable? Maybe 1 or 2% of the time
> but surely not for the vast majority...
On a case by case by basis, no of course not, unless you have some
prior agreement or implict trust - that's why we have social notions
like brand management and technological ones like spam filters. But
it depends on what you're doing with this data. if you're
aggregating large amounts it, you get some amount of smoothing -
Amazon reviews remain useful despite the fact people write them.
I've always felt that the stuff that Tim BL and Joshua Allen talk
about requires a "signal processing" layer to get from data to
metadata. That's not present in the SemWeb layer cakes I've seen.
Miles Sabin quite rightly mentioned that GOFAI has yet to prove
itself worthy, but there are systems out there where stuff
remsembling usable GOFAI has been built, but on top of scruffy AI
tech, not handcrafted ontologies.
> Actually the only success of meta-data publishing is about publishing table
> of contents.
So I just need to find one counter-example to refute that... ;)
> People have enough incentives to be honest about providing
> location, date and a short description about a resource.
I think there is a significant, but orthogonal, architectural issue
here. On the Internet, it's too hard to prove you're not a dog.
> The kind of meta-data Doctorow is speaking about is not about table of
> contents but more other kind of meta-data like for instance providing
> keywords or category for a resource.
Some of it is, some of it isn't.
> In that case I would trust more an
> automated source than human beings trying to extract as much cash as
> possible from our pocket. And even automated sources can be manipulated (we
> call that Search Engine Optimization) by the very same people.
No disagreement from me.
> Bottom line, I really don't know if the semantic web can go beyond the table
> of contents or some very limited applications.
Suppose it was 1935 and I said: bottom line I don't know whether
these proposed computing machines can solve the
Entscheidungsproblem. Does that make computing today not useful by
virtue of being restricted to limited applications (or as Picasso
pointed out, computers can only give you answers)? This stuff
doesn't have to be formally correct or all encompassing to be useful.