Skim, Not Aggregate
Posted on:: November 26th, 2003
I don’t want an aggregator. I want a skimmer.
Functionally, the two are quite alike. But while an aggregator pulls together the stuff I want to read, the point of a skimmer is to let me figure out what not to read.
I’m not looking to read automatic summaries because, well, they suck (see the next blog entry). Once I’ve decided to read something, the skimmer lets me read it in full. I just want help in knowing what not to read.
(As far as aggregators go, I’m continuing to like Bloglines.com.)
Dan Bricklin has long studied skimming. See the Good Documents site, e.g., this page from 1998 that includes the skimmable version of the Starr report.
Categories: Uncategorized dw
What we need is a Bayesian filter for an aggregator that will take in all the blog feeds, and pick out the ones that score highest in the interests we’ve demonstrated in the past. Now, THAT would be a cool toy!
How will that help you find what might interest you in the future?
Still have to put in the work thinking and looking.
Ted,
That’s exactly what I did and it works surprisingly well. And, since I’m not selling anything, I’ll plug it here.
Basically 348North News is a normal aggregator in much of the way you think of it. However, it allows me to identify keywords or themes that it puts together into phrases — and then matches up the phrases with like articles. Like a cross between Google News and Daypop (but that makes it sound much more complex than it is).
If you want to see an “interests” based summary for me, check out the Phrase Index. I use fairly general keywords so as not to miss out on the future items. (good point, Jon)
I think JonUdell has done some experimenting with running bayes against blog entries.
In terms of writing for skimmability, Robert Horn’s “Structured Writing” ideas are interesting…
http://webseitz.fluxent.com/wiki/StructuredWriting
Keeping up to Date With Your Reading
Apparently there’s this guy, living in New Jersey, age 79, who is really devoted to reading the newspaper. He doesn’t miss a day, literally.
Jon asked: “How will that help you find what might interest you in the future?”
It won’t — just as a Bayesian spam filter won’t flag a spam mail with different patterns than those of the past.
I don’t favor the Bayesian filter as the only tool to use – you still have to keep up with the industry. But if you’re specializing in FoxPro and SourceSafe and Linux and MySQL and WebConnect solutions today, having those articles with high scores can keep you up on what’s important currently.