[tti] Morning
Andrew took us through IBM’s Webfountain project, which crawls gigantic piles of data looking for meaning. He pointed especially to applications using it to look for the effect of marketing campaigns on Web pages, although it has other uses, of course.
Ian Black of Autonomy says that the Auutonomy project head at Ford’s training department says “Metadata is for the birds” because his department generates 5 million new objects per month, too much for manual tagging. Autonomy wants to provide systems that push info to users without requiring them to interact with it beforehand.
Eric Bonabeau of Icosystem says that we tend to believe that the future is present in the mass of data and we just need to find it. But we use existing patterns to search the data, which can’t turn up new patterns. Humans are amazing pattern-detecting machines but we’re terrible at exploring alternatives. Eric suggests selective breeding: make new combinations, look at the results, pick the most interesting results, recombine them, etc. He does a live demo that discovers a “hidden bagel” in a 50-dimensional financial data set. (I have no idea what that sentence means once it gets past the word “bagel.”) It helps not to know what you’re looking for. (He’s posted an online experiment interactive evolution.) [Disclosure: I did some consulting to Icosystem a couple of years ago.]
Categories: Uncategorized dw
It bugs me when people conflate “metadata” with “manually-entered metadata”. What do they think things like file size and last-modified dates are?
http://167.89.225.170/2005/02/09#a537
eContent mapping
[tti] Morning .
Metadata is for the birds
David Weinberger, writing from the TTI-Vanguard conference last week, quotes Ian Black of Autonomy as saying, “Metadata is for the birds.” I’m starting to see how this comes about.
Interesting posting, David. Metadata has been much on my mind lately. (See my recent posting)
In the case of Ford, do you know whether they’ve considered the folksonomy approach? I know it wouldn’t replace structured metadata, but it might be a good-enough supplement to structured metadata in some cases. (In other words, it might be better than nothing — and less tedious.)
Also, you wrote above: “Humans are amazing pattern-detecting machines but we’re terrible at exploring alternatives. Eric suggests selective breeding: make new combinations, look at the results, pick the most interesting results, recombine them, etc.”
…Regarding that, this may sound a little weird but in my own life I’ve found that using symbol-rich card sets (such as tarot or oracle cards) can help me think creatively about complex situations. What makes this technique/hobby so valuable to me is that it allows me to look at the symbols and consider “How might this relate to situation X?” Sometimes the result is gibberish, but surprisingly often this gets me thinking in new directions, or helps me see a context I hadn’t considered before. Randomness and symbols can be surprisingly powerful in both creativity and interpretation. (And no, this isn’t about fortunetelling, just thinking creatively.)
I wrote more about how randomness and symbols can aid creativity and might be incorporated into content management systems in this article.
By the way, I found your article via Jack Vinson’s blog.
Thanks very much. I enjoy your work.
– Amy Gahran
Editor, CONTENTIOUS
Manual Tagging – limitations
Manual Tagging: “Ian Black of Autonomy says that the Auutonomy project head at Ford’s training department says “Metadata is for the birds” because his department generates 5 million new objects per month, too much for manual tagging. Autonomy wants to…
Two additional time-proven methods of inducing variety and mutation into one’s creative gene-pool are composing poetry and consulting the I-Ching.
Handicraft and the personal touch
… probably just snotty rantings.
bonabeau made no sense at the conf and after i asked a question about finding info in data he gave me the dot-com answer of “just keep looking, its there” he offered no examples of where this helps. Verdict, hot air, i’ve invested in real companies with ideas about data-mining tech, and Ico ain’t even close