« Paul Otlet |
Main
| Selling Social Software Event in London »
November 20, 2003
Otlet: Some ideas die because they are wrong
Posted by Clay Shirky
Though I found
the Otlet article Ross posted fascinating, this passage
Distinguishing Otlet's vision from the Bush-Nelson (and Berners-Lee) model is the conviction—long since fallen out of favor—in the possibility of a universal subject classification working in concert with the mutable social forces of scholarship.
Otlet's vision suggests an intellectual cosmos illuminated both by objective classification and by the direct influence of readers and writers: a system simultaneously ordered and self-organizing, and endlessly re-configurable by the individual reader or writer.
requires a response.
The failure of "universal subject classification working in concert with the mutable forces of scholarship" didn't happen because that idea fell out of fashion -- it was fashionable as recently as 1998, with people being paid fabulous sums of money to pursue it. It failed because it _does not work._
Yahoo had professional ontologists on staff (there's a business card, huh?) to do just this -- work out a highly detailed set of relations between entities that was both hierarchical _and_ web like (analogous to a directory structure with symlinks) and they tried to rank the items within those systems based on relevance, itself a result of both the contents of the file and self-reported characteristics by the creators, in things like the META keywords list.
It was, in other words, "an intellectual cosmos illuminated both by objective classification and by the direct influence of readers and writers." And it sucked. Sucked sucked sucked. We didn't even know how bad it sucked until Google came along and (its hard to remember this even five years later) saved the Web from drowning in its own waste.
And Google did it by saying what every system that scales to internet size says: fuck ontology.
Google's intuition explains why both Otlet then and the Semantic Web now are doomed: classification shemes don't scale, because at billions of documents, the width vs depth tradeoff (many top level categories vs many deep trees) stops being a tradeoff anymore because systems become both too wide _and_ too deep. They also don't scale to include alternate worldviews that use alternate classifications.
Google instead followed the Jakob Nielsen dictum: don't listen to what the users say, for they are unreliable and will game the system; instead watch what they do, and derive meaning from that. And that worked, in a way that the Mundaneum never could.
Not knocking Otlet, mind, he was working in the tenor of the time, but the idea that the rock candy mountain of a universal global ontology didn't appear because of dictates of fashion is nonsense.
Comments (12)
+ TrackBacks (0) | Category: social software
- RELATED ENTRIES
- › Spolsky on Blog Comments: Scale matters
- › "The internet's output is data, but its product is freedom"
- › Andrew Keen: Rescuing 'Luddite' from the Luddites
- › knowledge access as a public good
- › viewing American class divisions through Facebook and MySpace
- › Gorman, redux: The Siren Song of the Internet
- › Mis-understanding Fred Wilson's 'Age and Entrepreneurship' argument
- › The Future Belongs to Those Who Take The Present For Granted: A return to Fred Wilson's "age question"
1. Liz Lawley on November 20, 2003 11:44 AM writes...
And yet...and yet...libraries--real brick-and-mortar libraries--linger on. As do the pay-per-use databases based on broadly accepted ontologies like LCSH, MeSH, et al. Librarians knew about--and internalized into their practices--ages ago what you attribute to Nielsen. Unfortunately, the library literature on "information-seeking behavior" is largely ignored by today's usability experts and information architects. :/
Carol Ou had a nice piece about libraries and metadata in response to your semantic web essay that touches on some of this.
Permalink to Comment2. Mercutio on November 20, 2003 11:47 AM writes...
Clay,
Did you read any of the responses to your last post on the Semantic Web? Try this one http://www.ftrain.com/ContraShirky.html
The Semantic Web is not about "universal ontology". It is about ontologies for many small domains, just like database schemes, with the added advantage that they can work together.
Yes, Google's approach is more successful than Yahoo's, but what about when Google can use whatever partial ontology is available to improve search results. That hybrid approach is the Semantic Web.
Permalink to CommentBTW You're making an ass of yourself with your rants.
3. Ross Mayfield on November 20, 2003 12:47 PM writes...
Clay, you have full right to ignore the anonymous ad hominem
Permalink to Comment4. Dave Winer on November 20, 2003 3:26 PM writes...
What about attributed ad hominems?? ;->
Permalink to Comment5. Ross Mayfield on November 20, 2003 4:43 PM writes...
Well, that's a matter of style Dave ;-)
Permalink to Comment6. Lindon Parker on November 20, 2003 4:49 PM writes...
...maybe persoanl ontologies may work better, try looking at this article http://www.xml.com/pub/a/2003/11/19/udell.html
Permalink to Comment7. Escalus on November 20, 2003 5:19 PM writes...
Hey Ross,
Not "ad hominem", friendly advice.
Clay has missed the fact that the Semantic Web is the infrastructure for distributed, incremental, messy, personal ontology - the same virtues he sees in social software.
The sooner he eats a little crow and stops flailing and swearing at his horse's carcass, the healthier he'll be.
M.
Permalink to Comment8. Tim Oren on November 21, 2003 9:06 PM writes...
A problem here is that the phrase 'Semantic Web' is being used to represent two different things. The first is a collection of mechanisms facilitating integration of data and programs across services in an open standards framework. The second is a grand vision of a global symbolic ontology. The first is useful, necessary, and likely inevitable. The second is just as much bullshit as when CYC tried it.
Permalink to Comment9. Chris Tolles on November 22, 2003 6:14 AM writes...
Clay:
Directories (and their underlying ontologies) provide different value - and Google very definitely uses an ontology and a directory to add to its search results - It uses the Open Directory Project, and its 460k categories (as well as providing its own version of the directory).
Not defending the "Semantic Web" per se, but pointing out that you're making a stronger case using Google than I think Google would purport. If you want a list of VC firms, or universities In New York, a directory is pretty nice, and a useful thing.
While AOL has certainly not done much lately to champion and support the ODP, a peer review community having built the directory structure that it has is something to put into the argument here.
Chris Tolles
Permalink to Commentco-founder, Open Directory Project
10. Lucas Fletcher on November 23, 2003 2:01 AM writes...
Here's an analogy, see if you can follow it before reading the explanation after:
text-based searches : web :: perl : programming languages
Okay, that wasn't so hard. Perl and other text-based languages are primarily oriented to handling unstrucured text data. As is Google. Maybe this suggests possible shortcomings of pure text based search, such as: IT'S A BIG UGLY HACK! (sorry for shouting) I honestly believe that the only way the web can evolve is for more and more data to become more and more structured. I agree the semantic web is wrong-headed. But ontology is a prime ingredient in any attempt at structuring the chaotic mess of malformed HTML documents written for way-too accomodating browser enginges.
Permalink to Comment11. Alex Wright on November 26, 2003 12:20 PM writes...
Since Mr. Otlet can't be here to defend himself, I suppose that task falls to me.
Alas, I fear Mr. Shirky may be pressing Otlet into service as a straw man for advancing his argument about the futility of universal classification schemes for the Web. Which is a fair argument, but not a fair use of Otlet.
To portray Otlet as a champion of authoritative classification is to misconstrue his legacy. If you want to set up arguments with nineteenth century proponents of universalist classification, better to go after the Victorian "true believers" like Dewey, Panizzi, or Cutter. Or, like Borges, you could even go back a few hundred years and mix it up with Roger Wilkens. But please let poor Mr. Otlet have his due.
Otlet matters not because he believed in universal classification, but because he recognized the importance of associative trails and collaborative authorship, directly presaging the visions of Bush, Nelson et al. His belief in the possibility of marrying classification with social context is important not because he advocated classification, but because he foresaw the possibilities of self-organizing systems.
I've written a more detailed response over here:
http://www.agwright.com/blog/archives/000794.html
Permalink to Comment12. Meme Engineer on December 30, 2003 1:24 PM writes...
It is true that the inherent rigidity of Paul Otlet's classification scheme would be a fundamental shortcoming in this modern age of modern computer hardware, but that doesn't detract from the man's brilliance in working with his time's available tools.
One may feel deeply sad for him personally.
Permalink to Comment