Commercializing the Semantic Web

October 22, 2003

Kendall Grant Clark

This week I am attending, along with nearly 500 other people, the 2nd annual International Semantic Web Conference (ISWC), on Sanibel Island, near Fort Myers, Florida. In addition to ISWC, there have been meetings of DAML principal investigators, of the various communities comprising the Semantic Web Services Initiative, meetings of the Ontoweb organization, and a few other things which haven't quite made it on my radar screen.

The Dot-Bomb Disruption is Not Another AI Winter

Though I haven't heard anyone say this explicitly, it's important to say that the dot-bomb disruption is significantly different than the Artificial Intelligence Winter, that sustained funding glut which, at least in some people's view, killed AI and Common Lisp as thriving commercial endeavors. Whether people deny it or not -- and most of the denials seem to be motivated by AI Winter recurrence fears -- the Semantic Web really is an attempt to reconceptualize and reengineer AI for the Web.

Okay, perhaps that's overdrawn, since there are particular subfields of AI -- logic programming, knowledge representation, and, to a lesser extent, machine learning -- that are especially relevant to the Semantic Web. But let's not beat around the bush: the academic fields from which Semantic Web researchers are overwhelmingly drawn are AI fields. And many of the core techniques, models, methods -- including knowledge representation, ontology-based reasoning, description logics, and so on -- of the Semantic Web are AI techniques, models, methods.

Why is this a useful or salient distinction? Because, first, there is an enormous difference between attempting to webize AI and attempting to AI-ize the Web. Most of the fears about the Semantic Web from the web-hacker community, whom academic researchers ignore at their peril, are that it is an attempt to AI-ize the Web. Such fears, at least in my view, and I try to keep toes in both communities for precisely these reasons, are almost wholly empty. Second, the dot-bomb disruption has had a range of effects on the Semantic Web, but that range is different than the effects of the AI Winter.

Network Inference: A Semantic Web Company

One of the threads of ISWC I have been trying to disentangle is the question of how Semantic Web technologies will be industrialized. One of the shining stars in this regard is Network Inference (there are other Semantic Web companies, but none of them seems as far along).

Network Inference is a full-on Semantic Web company, and it's solving real problems for real clients. These two points, and particularly the latter, are important and to some extent surprising. While I tend to follow this area closely, I am frankly shocked that any bona fide Semantic Web company has real clients already. It seems very early for that kind of success, though whether or not it translates into long-term business success is an entirely different question.

Is Network Inference really a Semantic Web company? Yes, it is: if there are any technologies which are undoubtedly Semantic Web technologies, they are RDF and OWL (for more about OWL, see my recent column "The Semantic Web is Closer Than You Think"). Network Inference's technology is centered on Cerebra, its own world-class OWL (and thus, Description Logic) reasoning engine.

Network Inference uses Cerebra in a variety of scenarios, including as the main driver of policy management applications, as well as a component embedded into larger application suites. As a way of pushing the transition from non-semantic to semantic technologies, Network Inference deliberately targets very complex problems and systems. In addition to OWL, they are using XQuery over SOAP as a way to do front-end, back-end integration between Cerebra and application and web servers. I expect that Network Inference will eventually move away from XQuery, at least if the W3C can get an RDF query working group off the ground.

Just Don't Call It the "Semantic Web"

For reasons I don't entirely understand, the term "Semantic Web" tanks with corporate clients, with venture capitalists, and in other non-academic contexts. This may yet be a hangover from the AI Winter, but the interesting difference is that, as I discuss below, the reaction is mostly to the label and to its perceived implications, rather than to the technology itself. "Web Services" does much better, and one of the things Network Inference seems to have done, at least at the marketing level, is to hitch its semantic wagon to the web services star. (This is a move I suggested, though more in the research than marketing context, in an article last summer, "The True Meaning of Service".)

Perhaps the most obvious corporate pitch for the Semantic Web is as a kind of clever enterprise application integration solution; but, as the Network Inference folks like to point out, the EAI space is very competitive and very saturated. Rather than competing with the EAI heavyweights, Network Inference is attempting to compete in a slightly different area, while trying at the same time to define a new space altogether. Instead of pitching its technology as an EAI solution, Network Inference is presenting it as a kind of policy management solution; that is, as a decision service which runs over top of the integration layer. The value of such a service is located mostly in the maintenance part of the application lifecycle.

Given the problems with the various application spaces, Network Inference has apparently been working to define a new application space, one which the Gartner Group has coined as "semantic oriented business applications". That doesn't raise the hackles that "Semantic Web" raises; it's different than EAI, and it's nicely distinguished from "Web Services".

A Big Win?

In an era of shrunken and shrinking IT expenditures, recouping costs in application and project maintenance phases is particularly important and useful. I suspect that there is also a psychological effect as well. After all, spending money on maintenance of existing systems just feels different, both for individuals and organizations, than spending money to acquire new stuff (that is, capital expenditures); the former feels more like wastage spending than the latter. Technologies that reduce maintenance costs are winning technologies over time. Here we see one of the differences between the dot-bomb disruption and the AI Winter.

In the AI Winter case there seems to have been a general souring on the technology itself; people, whether justifiably or not, turned away from the technology and funding organizations turned away from or attenuated investment in the ongoing development of the technology. But in the former case, which we're in now, people and organizations haven't so much soured on web technology as they have simply run out of money. IT budgets are decimated largely because corporate budgets are decimated, largely because the global economy sucks and has been sucking for a few years.

This makes a difference in the way that the Semantic Web can survive the dot-bomb disruption. The public funding of Semantic Web research, while not flush or excessive, is robust enough that research continues along at a lively pace, as witnessed by the fact that ISWC attendance is much higher than was anticipated. This growth makes for uneasy laptop access to conference power outlets but is a good sign overall.