xml-dev - RE: RDF, the "semantic web", and the nadir of AI (was RE: Realist ic pro

[Jonathan.Robie@SoftwareAG-USA.com]

Title: RE: RDF, the "semantic web", and the nadir of AI (was RE: Realistic p roposals to the W3C?)

Mike.Champion wrote:

> it's hard to shake the feeling that the 'semantic web' is 1980's
> AI hype on life support. Could anyone set me straight if this
> is terribly unfair?

I don't know whether or not the semantic web will succeed, but the idea does make some sense.

The problem with semantic networks in the 1970s was not that it is hard to reason about databases of assertions. The problem was that it was not in anybody's interest to type in all the assertions you would like to be able to base your reasoning on. The web has a huge amount of data in machine readable form, but a similarly large database of assertions does not yet exist. If you can come up with a business model or some kind of ego incentive to make people supply the missing networks of assertions, then software can use the network of assertions an a number of useful ways.

A second  problem with RDF is the lack of programming and query interfaces to allow you to marry it to databases, XML repositories, programs that generate information in various ways, etc. Since a lot of information on the web is dynamically generated, this is significant.

A third problem has been that topic maps, RDF, and XLink have overlapping functionality. If the three come together in a useful, synergistic manner, a lot of things just might start popping. Suppose topic maps build on the architecture of RDF, and RDF uses the superior serialization formats of topic maps, which abandon HyTyime links and adopt XLink for their serialization. I'm not up on the current status, but I know people are talking about these kinds of things in various settings. If that all happened, we would have three small, creative camps working together, and it just might achieve critical mass.

Also, I don't think the term "semantic" is necessarily equivalent to "voodoo". Each namespace identifier can identify the language in which its data is encoded, and what is semantic for the web as a whole may be syntactic in the language associated with a namespace. I think that's what Tim meant when he said:

<Quote src="http://www.xmlhack.com/read.php?item=527">

   the crucial thing is to recognize that the namespace
   identifier identifies the language of the message and
   so indirectly its meaning.

</Quote>

A key word in the above quote is "indirectly". The meaning can only be understood if the language conveys it in a way that is either (1) syntactic to some program, or (2) understandable to a human to whom instances in the language are presented.

Jonathan