Friday, January 26, 2007

A Great Day for Specificity

dbpedia.org - Using Wikipedia as a Web Database
[...] dbpedia can also be seen as a huge ontology that assigns URIs to plenty of concepts and backs these URIs with with dereferencable RDF descriptions.
We have advanced a tech level!

This is really good timing, as I just recently considered using TDB URNs for referencing non-retrievable things and concepts (using TDB-versions of URLs to Wikipedia articles and web sites). Finding out if the idea of TDB and DURI URNs have been long since abandoned was my next step down that path. (Then there's the use of owl:InverseFunctionalProperty and bnodes (or "just-in-time"-invented ad-hoc URIs) and throwing in owl:sameAs if official ones are discovered..)

With DBPedia, a lot of that becomes unnecessary. The availibility of public, durable URIs for common concepts will surely ease the integration of disparate sources of knowledge. That is, if we start to use dbpedia-URIs in our statements.

And gone will be the strangeness of saying "I'm interested in [the word 'semantics']", "This text was edited with [the website <http://vim.org>]" and "I was born in [the wikipedia article about 'Stockholm']"..

2 comments:

Larry Masinter said...

You might have missed the point of 'tdb':

http://dpedia.com/anything

is still the URI of the web page at dpedia. If you want a URI that identifies the thing described by that web page and not the web page itself, you need some other way of 'grounding' the abstraction in the real world.

Some people want to usurp the fragment identifier '#' to do that grounding, but '#' already has a meaning for subpart referencing.

So, to distinguish between me and my web page, you either need some other external mechanism ('the thing described by my web page') or you need a new URI scheme (urn:tdb or tdb).

Niklas said...

Ah yes, in general most URI:s do point to regular pages and thus identify those in themselves. For all of that I definitely see a good use of TDB. But dbpedia themselves delcare (the bulk of) their URI:s as concept identifiers (those under <http://dbpedia.org/resource/>):

"[...] dbpedia can also be seen as a huge ontology that assigns URIs to plenty of concepts and backs these URIs with with dereferencable RDF descriptions."

That's "good enough" by my standards. It may still be an open (albeit mainly philosophical) question about the "correctness" of that though. And I definitely see it as relevant to point out the misnomer in using URI:s to represent "real things" carelessly ("I am not my email").

Do you consider the dbpedia approach (also taken by e.g. Semantic MediaWiki, see e.g. Cool URIs for the Semantic Web) to be conceptually flawed? If so, could it be corrected if there was an owl:sameAs statement for each relevant URI:s pointing to an "urn:tdb"-based one, or do you find the scheme of a URI significant for the nature of the thing it identifies?