Primed for the Semantic Web

November 8, 2000

Leigh Dodds

Prompted by Edd Dumbill's "The Semantic Web: A Primer" article published last week on XML.com, the RDF Interest Group have been discussing their own views of the Semantic Web. This week's column summarizes the discussion and reports on other progress in the RDF arena.

Vision

Noting that Dumbill's primer article is a useful starting point, Dan Brickley invited others to share their vision of the Semantic Web.

This strikes me as a useful article, both in terms of providing a discussion strawman on what 'Semantic Web' might mean in practical terms, as well as for surveying some key technologies, making the point that XHTML, XSLT and mainstream XML apps will be critical tools for the SW.

Rather than gabble on about what I think Semantic Web might mean I'm pretty curious to hear what folk on this list understand by the phrase...

Outlining his own viewpoint, Brickley believed that the Semantic Web is the Web as it was originally conceived.

... the Semantic Web will be the Web - we're not talking about a 'new' Web here, just finishing off the job with the existing one.

With this in mind, Brickley encouraged others to research Web history, particularly the archived design discussions on the W3C www-talk mailing list.

I encourage anyone interested in Future of the Web ("Semantic Web" or whatever you prefer to call it) to spend some time digging into Web history. Both the original proposal and the www-talk archives are worth spending some time reading.

...we need to keep Semantic Web discussions grounded in a sense of history and evolving Web architecture...

The Semantic Web then is not something that will be built from scratch: it's the goal many have been pursuing for the last ten years. Brickley discusses this further in "Nodes and Arcs 1989-1999, The WWW Proposal and RDF: Then and Now"

Responding to Brickley's call for discussion, Graham Klyne described his vision of the Semantic Web as an "end-to-end architecture for content".

The semantic web aims to do for information what the Internet and WWW have done for data. Today, the Internet, and the WWW in particular, allow data to be exchanged between an arbitrary pair of systems, without regard for or discrimination among different system types, data content or the service to which the data relates.

Today, information tends to be bound up in the data format of the application for which it is designed. Without the right application, the information content of data is inaccessible. The semantic web will make the information content of exchanged data accessible to any application that understands semantic web protocols.

Bill deHora thought that the Semantic Web is about "significance," something that the Web is currently lacking.

The big idea behind the Semantic Web is that we can do more significant stuff than turn some words blue and draw lines underneath them. We can for example, turn some words blue and draw lines underneath them and then say "what's on the end of this link is written by the same person as this is". Link significance makes the difference and significance is in the eye of the beholder. In other words the current Semantic Web isn't very rich in significances. Richness was traded for Acceptance, a very useful deployment pattern, but you borrow some trouble nonetheless.

Echoing deHora's final comment, Matt Jensen drew a parallel between descriptions of the Semantic Web and the flaws in our current situation. Jensen noted that the successes of the Web were realized by sacrificing the scope of what was trying to be achieved.

There were many people in the 80s working on hypermedia systems, and a significant reason that they stalled and the WWW took off is that they cared about ensuring consistency, bidirectional links, etc., and Tim was willing to let go of that. The result is >1 billion WWW pages, and probably >10 billion links. A small percentage of the pages are broken, but on the whole the WWW provides tremendous value.

Similarly, I view most of what has been done in AI as focused on consistency, correctness, etc., which (so far) has limited the successes it can claim. If you're looking for a Semantic Web that can give you "truth", we've got a long wait. If you're looking for something that improves search results through related concepts and simple inferences, in a few years you should be able to get something that's useful, but not perfect.

Seth Russell also commented on the importance of acceptance.

There is a trade off between precision of expression and growth and acceptance. The more precise and logically consistent we insist upon RDF validations, the more difficult we make implementations, and the less likely anyone will use them. This thing should unfold like an onion, with the outer shells being super simple and robust and the inner shells being rich and complex. But we'll never survive long enough to get to that rich inner core unless the outer shells gain wide acceptance.

The interesting aspect to this discussion is the positioning of the Semantic Web as an evolutionary progression rather than alternative architecture. It shows signs that the RDF community is keen to begin identifying areas for development, converting the lofty aims into realizable milestones. Projects such as the DAML Program are amongst the first to start tackling the Semantic Web head-on.

Momentum

As Dumbill's primer noted, there are existing tools that can be used to begin developing the Semantic Web. But the RDF community is not letting the grass grow under its feet: more tools are appearing all the time.

In the past week a new version of Redland, the RDF Application Framework, has been released, along with an early version of RDFStore. RDFStore is a series of Perl modules used for manipulating RDF databases. Of course, if you have an RDF database, then you'll be needing an RDF query language. Enter RQL, which "propose[s] a new data model and a query language for RDF descriptions and schemas."

The RQL effort seems initially targeted at supporting the requirements of the many Community Portals which are springing up across the web. In fact, even the Semantic Web now has its own community site: SemanticWeb.org. The introductory pages note that:

[w]ork is going on to realize tools and techniques, which will help to create the Semantic Web. This site is dedicated to collect these approaches, to explain them and to be a forum for people interested in the Semantic Web.

And there is now even a Semantic Web Journal that's concerned with

modeling semantics of web information, and covers theory, methods, and applications.

The surge of interest in RDF which the Deviant has noted over the last few months appears to be continuing with some of that momentum being transferred to Semantic Web efforts.