Edd Dumbill reacts passionately to critics of grassroots-driven ontology creation (a representation of terms and their interrelationships) creation such as the Semantic Web as promoted by Tim Berners-Lee. He writes:

Photo credit: Annette Gulick

"Both Shirky and Udell seem to be pretty much convinced the Semantic Web requires, from the outset, globally agreed ontologies. It seems more that they've set up a straw man.

I had always envisaged that in the same way user interface and other conventions have emerged from the messy web, so would ontological
conventions. Messy, but good enough."

Stephen Downes adds: "I will say, if you are building ontologies now, prior to use, you are probably making an error. Same thing if you're creating canonical vocabularies.

Categorization is good (reason is impossible without it) but categorization belongs to the language, not the librarians."

But Jon Udell sets off some reasonable alarms when he writes: "Semantic-Web naysayers think people and organizations can't be bothered to assert machine-readable facts about themselves. And, today, that is undoubtedly true.

But when others assert facts about you -- as they increasingly will -- the tide could begin to turn. Individual acts of self-defense may ultimately combine to bootstrap the semantic Web."

Could it be then, that a better balance between these two forces may be achieved by leveraging a bit of both?

If we provide initial, open-ended ontologies that users can add and modify as well, don't we provide an easier path for developing sensible initial categories?

Couldn't this facilitate a faster and more focussed user-centered ontology development?

And then, once effective ontologies emerge from user-driven preferences couldn't we extract and identify key categories and sub-topics as defined by the users and then utilize those as a new starting plateau for further refinement?

What do you think?