I didn’t expect anyone to pay much attention to my
href="http://www.oreillynet.com/pub/wlg/7823">modest efforts to
mashup “Web 2.0″ and “Semantic Web”. I guess I didn’t expect much
feedback because I’ve always thought of the Semantic Web and Web 2.0
as somehow connected or related.
And I’ve always thought that because I know what some people don’t
know: there isn’t one Semantic Web, there are two of them.
There is the scruffy Semantic Web, and there is the
neat Semantic Web — these terms come from two different
tendencies (we won’t call them factions just yet) in the Artificial
Intelligence part of computer science, the Scruffies and the Neats.
As I said last
time, the Semantic Web can be thought of as an effort to
reimagine, and to rebuild, part of AI, the part called Knowledge
Representation, as having the Web as its proper environment. In other
words, the Semantic Web is webized knowledge representation or
knowledge representation added to the Web.
My point in the last weblog entry was that it may best be added to
the thing that’s starting to be called Web 2.0. And, yes, Virginia,
“Web 2.0″ is a marketing term. So what? Language itself is a marketing
term; it’s all just scribbles and noises in patterns intended to help
someone get closer to achieving one of their projects.
The funny thing about, say,
Shirky’s anti-SemWeb schtick is that it more or less recapitulates
very old debates in AI itself, all while adopting a very mannered,
very anti-AI pose.
The problem with his schtick, if I may be so bold, is that either
he doesn’t really understand AI or KR very well or his attempts to
popularize it are very poor. All of that talk about syllogisms and
ontologies is really just rubbish. It’s roughly equivalent to someone
writing a very popular article criticizing the contemporary practice
of physics as being all about the href="http://en.wikipedia.org/wiki/Phlogiston_theory">phlogiston. The
point isn’t that Shirky is wrong to claim that modern logic (since,
say, Frege) is about “syllogisms”, but that he’s laughably,
embarassingly wrong. OWL ontologies have about as much to do with
syllogistic logic as href="http://en.wikipedia.org/wiki/M-theory_(simplified)">Edward Witten’s
M-theory has to do with phlogiston or the ether.
Setting Shirky aside, all the talk amongst
advocates like Robert McManus about getting by with “just enough”
semantics and “good enough” standards is exactly the kind of debate
that Scruffies and Neats have been having for ages. As my friend and colleague Bijan Parsia likes to say, the Scruffies just
are scruffy precisely because they think less semantics more widely
realized will work; the Neats are neat precisely because they think
you really have to get the semantics and the formalisms and the logics
just right before getting a lot of benefit.
See, Web 2.0 and Semantic Web folks really aren’t that far
Who’s right, the Scruffies or the Neats? I don’t know who’s right, and, more,
I don’t care. Both are doing their thing, and as with
all evolutionary struggles, reality will be the harshest and best
judge. Because of my background in the kind of philosophy that always
made my brain hum the happiest (Rortyean pragmatism), I’m happiest
trying to subvert such distinctions rather than trying to inhabit them. Why choose till or unless you absolutely have to?
But the Scruffy Semantic Web and Web 2.0 are a good fit, not only
because their advocates care about the same things, but also because
their Good Enough intuitions are congruent.
For what it’s worth, the key Semantic Web technologies, RDF and
OWL, show signs of the struggle between Scruffies and Neats. (I’ll
explain some of the details about that claim in the comments if
anyone’s really interested.) I also think Tim Berners-Lee, who’s
original vision for the Web just was the Semantic Web, is a
scruffy. Which makes him a natural ally, at least in my opinion, of
Web 2.0 advocates. Of course Sir Tim speaks for himself!
Some people call Web 2.0 the Web as a platform. I don’t think
that’s correct; or, at least, if it is, it’s not very interesting. The
plain old Web is a platform — that’s the point of REST.
Web 2.0 as a marketing term is a shorthand term for a set of sites
and a set of informal policies and attitudes that have coherent,
stable properties. I think every Web 2.0 site is both a web app and a
web service, and they’re almost all lightweight REST web services
(instead of heavyweight SOA services). They all use semistructured
data (XML, RDF, JSON, YAML, etc.) to exchange representations of first
class resources. They’re all intended, and achieve to varying degrees
of success, to be used in a kind of mashup, mixup fashion. Which is to
say that the services are meant to be composable in some sense. (But
not like Semantic Web Services, where the composability is done by
computer programs using AI planning techniques and very complex OWL
ontologies, rather than by humans, by hand.) And every Web 2.0 site
has some kind of explicit or implicit commitment to letting you retain
control of your data.
I claim that every one of those properties is also something that
SemWeb folks have been working toward for as long as there have been
SemWeb folks. (Want to control yr data? The raison d’être
of RDF is to let anyone say anything about anything. Now
that’s control!) That doesn’t make SemWeb better or worse or
more important than Web 2.0. It makes them congruent. And congruence
is a precursor of convergence. Woohoo!
Finally, I believe every Web 2.0 site (or, as I’m going to start
calling them “apps/services”) exists firmly within the Scruffy side of
the camp: tags, for example, are all about the hunch (or is it a
fact?) that a little semantics (a very little, actually) being widely
distributed goes a long way.
So, thanks for all the feedback and enlightenment, but I haven’t
heard anything yet to put me off my convergence oats: Web 2.0 and the
Semantic Web are closer than you think, and SPARQL may very well be a
Don’t let me stop you…