The Semantic Web gets real

Widely regarded as science fiction, the Semantic Web promises a world where machines really understand each other. Eric Knorr asks if we've suddenly moved a step closer?

Depending on who you talk to, the Semantic Web is either a huge distraction or the future of the Internet.

Of course, there's no reason it can't be both.

Concentration on the Semantic Web to the exclusion of other efforts cost the W3C its leadership position in Web services (a position that has now defaulted to Oasis). And yet the Semantic Web may ultimately hold the answer to many of the knotty problems raised by Web services and other machine-to-machine communications schemes. A handful of major players including Adobe, Hewlett-Packard, IBM, Nokia, and even the Department of Defense have thrown resources at it--the Defense Advanced Research Projects Agency (DARPA), which founded the Internet, is a backer.

Surprisingly, I've even encountered one startup, Celcorp, that is putting Semantic Web technology to practical use in enterprises right now.

But first, let's review what the Semantic Web is about. Tim Berners-Lee, inventor of the Web and current head of the W3C, first cooked up the notion of the Semantic Web in 1998, which he described as "a web of data, in some ways like a global database." The basic idea is a Web in which remote machines can converse with each other in a meaningful way, rather than a Web limited to humans requesting HTML pages. Sounds a little like Web services, doesn't it?

But there's a big difference. Web services and XML provide a stack of protocols for machine-to-machine communications, but typically, this works only when a human being has determined the nature of a Web service. For Web services to go out and dynamically find other Web services, determine what those services do, and decide how they'll work together requires more sophistication than the current Web services protocols provide.

The Semantic Web enables such interaction by creating a metadata framework on top of the existing Web, where documents and services can be tagged with machine-readable descriptors that can't be misunderstood ("stock" in the financial sense could never be confused with "broth," for example).The bum rap on the Semantic Web has been too much sophistication--high concepts and arcane terminology that make understanding Web services seem like a walk in the park (for one of the clearest, briefest explanations out there, try Aaron Swartz's The Semantic Web in Breadth). And for the Semantic Web to work across the Internet, so-called "ontologies" (documents containing rules that describe how related data descriptors can be used) will need to be developed and deployed for a multitude of vertical vocabularies.

An oft-cited benefit of the Semantic Web is a world where intelligent agents take English-like instructions from users ("a week in Miami in March with a car for the lowest price possible") and cruise the Internet reading metadata to produce optimum results. But doubters say that the monstrous overhead of ontology development alone could knock the Semantic Web into the dustbin of nice-but-impractical ideas.

Celcorp is bucking this skepticism and actually using Semantic Web technology today. Apparently, the company even has corporate customers, including Liberty Mutual and Holiday Inn. With its Celware product, Celcorp helps businesses create meta-applications that draw upon (and sit on top of) an enterprise's existing application infrastructure, almost like highly sophisticated macros. Using a "semantic learner," Celware records and decomposes business processes into discrete activites. It then uses a reasoning engine to automatically assemble activities into business processes designed to meet such abstract business objectives as increasing productivity. In the end, an actual, executable Semantic Web application is generated that draws upon the functionality of existing applications at runtime.

"We fundamentally believe that the Semantic Web is the only way we can solve the very difficult issues of the growing complexity of information systems," says Celcorp's CEO Jas Dhillon. He acknowledges that building ontologies can be "very time consuming and painful." But, claims Dhillon, "one of our breakthroughs is that we can automatically extract the ontology of the system and the process" as the first step in developing its Celware applications. Another milestone in easing ontology development was the W3C's release last summer of a working draft for a Web Ontology Language (abbreviated OWL).

The entire computer industry now agrees that standards must continue to be developed that foster pervasive machine-to-machine communication across the Internet. Already, the stack of Web services security and business process protocols threatens to get so tall, that B2B Web services could fall into the same trap as CORBA, which promised distributed computing but ended up too unwieldy for most IT shops to implement. Put a semantic "wrapper" around Web services, however, and you could potentially reduce the complexity of what needs to be specified in protocols and XML schema--so that the machines themselves can shoulder much of the low-level, find-and-bind burden.

Realizing this vision requires widespread ontology development that few hard-pressed IT folk have the luxury to indulge in. As a result, It's perfectly conceivable that the Semantic Web could remain stuck on the drawing board forever--best case, the fruits are at least five years away.

On the other hand, given the history of the Internet, the DARPA endorsement is not trivial. As practical applications and easier development tools like Celcorp's arrive, enterprises may find themselves wading into the Semantic Web whether they know it or not.

Are Web services all the science fiction you need? Or does the Semantic Web concept appeal to you enough to actually explore it? E-mail Eric or TalkBack below.