Saturday, September 11, 2004

Making the Semantic Web Viral

The Semantic Web has not undergone the explosion of users that caused the early World Wide Web to be successful. Why? The answer appears to be that individual Web authors do not have simple tools for the creation of semantic linkages between existing content. The barrier to entry is currently substantially higher than it was in the early Web.

These tools cannot yet be built because two preconditions do not yet exist; a means to link Web resources while retaining meaning about the intended relationships between resources and a means of navigating (via URL) between all types of Semantic Web resources, both concrete (Web) and abstract (RDF).

I think there are ways to solve both problems with existing technology. Part of the answer, though, lies in the standardization process.

The way I see it, Web links with semantics need an attribute on a hyperlink that defines the relationship between the resources. The XHTML 2.0 Working Draft adds a 'rel' attribute to links to accomplish this. Thanks to Mark Birbeck for proposing this very necessary extension. The draft is in Last Call, so anyone who wants to comment should get on it.

Short of waiting for XHTML 2.0 browser support, one could always fake the intended behavior with Javascript and a little imagination. Pick up the 'rel' attribute contents and pass them along to a server that knows what to do about it. I hope to have some time to try this so I can flesh out a paper for WWW 2005. It would be cool to go to Japan, since I haven't been there in many moons.

Once you have links with semantics, you need to provide a means of navigating from the existing Web to the more abstract space of RDF and back again. You can already link a Web page to another Web page, but what do you do when you want to link to a person? Representing people as email addresses or even FOAF files is weak, silly and just plain insufficient. People exist only in Meat Space and so have to be represented in Semantic Web Space by some non-trivial means. RDF & OWL already provide a great answer for this: A person object can exist that link to email addresses, FOAF and all the rest. Now we only need a way to navigate it via the Web.

There is more than one way to do that. First, one could point to a node in an RDF datastore by URI and a transformation (e.g. an RDF query, some XSLT and a stylesheet) could present a view of RDF Space as a Web resource (an XHTML page, for example, but this could take many forms). The same technique (minus the RDF database) could be used if the RDF is serialized in an RDF/XML document or is extractable from a document (e.g. via GRDDL.

It would also be cool to return RDF information into a Web page without generating a new page, such as a layer box that lists several pieces of information about a link. Back to Javascrip to go get the data from RDF Space and present it. This would provide multiple destinations for a given link! I think that is pretty cool.

No comments:

Post a Comment