Feed on
Posts
Comments

W3C Semantic Web = Global Ontology after all?

Dec 12th, 2007 by Phil Dawes

I only just read Jim Hendler’s piece from last month “shirkying my responsibility”, in which he states that the W3C Semantic Web vision was never about a global shared ontology at all:

“Get it - we are opposing the idea of everyone sharing common concepts.”

This seems odd to me, because if that is the case and all communication on the semantic web is local then why is the basic system of identity the URI, a global identifier scheme?

On the contrary, I suspect that the W3C Semantic Web is predicated on global agreement: that all RDF documents containing a URI should use it to identify the same concept, otherwise the whole RDF inference stack breaks. A global ontology that’s defined in lots of inter-connected pieces scattered around the web is still a global ontology.

Posted in Semantic Web, web | Comments

Viewing 5 Comments

    • ^
    • v
    Don't think the two opinions conflict: "all RDF documents containing a URI should use it to identify the same concept" != "the same concept should be identified by the same URI in different RDF documents"...
    • ^
    • v
    That's true, but for semantic interoperability the same URIs need to feature in the communication somewhere - whether directly or via relationships defined in owl.
    • ^
    • v
    I don't get your point, Phil. As long as "global ontology" != "a single, predefined, giant ontology" [1]. Everything should be all right, or not? :-)

    [1] - http://www.w3.org/2001/sw/SW-FAQ#whgiantont
    • ^
    • v
    During the DARPA DAML project, which funded MIT's Semantic Web efforts at the time, Tim Berners-Lee envisioned the RDF web to unfold in a manner similar to the HTML web. Each development group would create an ontology (RDF taxonomy) independent of other groups. And that situation came to pass. A foaf:Person means the same thing as a cyc:Person but they have different URIs. The Linked Data initiative seeks to map these concepts together from various published ontologies, thus advancing the Global Ontology. For my own work, I am using the comprehensive OpenCyc ontology and extending it where needed.
    -Steve
    texai.org
    • ^
    • v
    Phil - I think you're missing the point - think of it as one namespace for all those URIs vs. many namespaces -- that is, in one case we'd all have to agree that "http://ex.com/foo#fish" was a particular thing with particular properties and etc, vs. allowing different people to have different ideas about what "fish" are - by keeping the info at different URIs and linking them together. So someone can define a vocabulary with some set of properties, someone else who disagrees, can do another, concepts can be linked or not, etc. In the "We all agree to one thing and sing kumbaya" ontology view, we must get all sorts of people from different cultures and approaches and views and religions and and and to all agree to everything - seems pretty daunting, right? On the other hand, the Semantic Web (i.e. web of semantic concepts) let's many different ones coexist, one can extend another, they can disagree, they can be simple or complex as needed, and they, most important of all, because of the URIs can be linked into a global whole -- but that global whole is not going to be consistent, carefully organized, and agreed to by everyone - and that's the key point we're trying to make. Note that this is related, but somewhat different, than what Steve Reed said - the linking of common terms is a piece of it, but the ability to have local agreement and real differences, that's not traditional to the knowledge approach - and requires new technologies and approaches, which is where RDFS and OWL come in...
close Reblog this comment
Powered by Disqus · Learn more
blog comments powered by Disqus