Archive for August, 2010

SwiftRiver sorts realtime social media streams

We are drowning in information, while starving for wisdom. The world henceforth will be run by synthesizers, people able to put together the right information at the right time, think critically about it and make important choices wisely.

E.O. Wilson, Consilience: The Unity of Knowledge

How can we evaluate citizen media for breaking news? Hurricanes? Earthquakes? Contested elections? Crowdsourced crisis information produces a flood of data. Individuals send emails, write blog posts, text friends, post to Twitter. Popular tweets are echoed and amplified. Eyewitness reports may be reliable, or mistaken – or planted disinformation.

SwiftRiver, an initiative of the Ushahidi project, aims to help humans aggregate and evaluate streams of social media. SwiftRiver is an open source platform for managing realtime data streams. Its services can be combined in different ways to serve the needs of crisis responders, journalists, and so on.

The first such app is Sweeper (now in beta), which intelligently filters data feeds for volunteers who then validate and geolocate the information.

Jon Goslin, TED Fellow and SwiftRiver director, talks about the five services underlying the apps.

  • Natural language processing to extract meaning.
  • Location context: considering how local the source is to reported events.
  • Reducing duplicates, especially those from Twitter.
  • Accounting for popularity separately from accuracy.
  • Reputation management: authority accruing to those for those who have a history of valued posts.

August 31, 2010 at 3:45 pm Leave a comment

From the Working Smarter Fieldbook

This weekend I received the Working Smarter Fieldbook (June 2010 edition) from the group mind at the Internet Time Alliance.

So right now let me get my bookmonger’s gripe out of the way: It’s a print on demand title from, so the fit and finish are ever so slightly rough. Cropping on the cover? Just a hair off. Graphics pasted in from presentation slides.

Fortunately, this isn’t an archival item. It’s a consumable. An “unbook,” never finished, in the authors’ terms. A convenient way of carrying around the-current-state-of-their-thinking, and frequently updated.

So I won’t feel bad about writing in it and dog-earing the pages. A few accumulated tea stains? Not a problem.

(And in that same spirit I’m practicing writing more, smaller blog posts. Working on speed rather than length.)

Today’s bibliomantic gem: Performance Support Trumps Training Every Time. (p. 153)

Post-training drop-off argues for ongoing performance support

This chart (via Harold Stolovitch and Erica Keeps) tells the story. During the training event, performance dips slightly as new ways of doing tasks are learned. After training performance is elevated, but forgetting begins.

Sure, that first post-training sleep helps to consolidate new memories, but other factors swamp the beneficial effect. Intervening tasks: email, phone calls, meetings. Lack of guided practice: the trainer’s off to another assignment. Change in context: how does this work in my job, really?

Here’s Working Smarter‘s bold claim:

An even better (and certainly cheaper) option is simply to cut out the training altogether and replace it with a support environment from the start.

And then they make the case for performance support systems.

Hm. Not convinced of the strong claim.

PSSs are valuable, certainly. I am reminded of early work on “training wheels” computer system designs, where learners begin with subsets of features. Progressive disclosure? Sure. Ongoing support? Sure. Total replacement of training with such systems? I think it’d depend on the domain.

My friend Molly just fininshed years of study to become a radiologic technologist. There’s two years of prerequisites before the program proper begins. Then, during the program, students move gradually from full time classwork to nearly full time clinical rotation. The volume(s!) of anatomy, physiology, and physics students have to learn… how would anyone approach it without classroom hours?

But Working Smarter has some of it right: that classroom knowledge can’t go into cold storage. It’s got to be woven in with application, and generalized across environments as students move from clinic to clinic during their training.

I’m an adult second language learner. I can’t imagine learning Chinese without classroom training, workbook practice, and audio files of simple conversations. I have less plasticity, but more metacognitive capability, than a young child immersed in a Chinese-language environment.

But ultimately I do have to take it out into the world. And it would be wonderful if the world provided some helpful post classroom coaching, rather than humiliation (or sheer incomprehension) when my pronunciation is off.

ADDENDUM: I find strong “unlearning” and “untraining” statements tremendously helpful. They reset my thinking about what learning environments can be. They keep me from reaching for the same tools all the time. So, colleagues pushing for more engagement, more improvisation – I’m looking at you, @webojunk – thank you.

August 22, 2010 at 3:39 pm 2 comments

Your world, with subtitles

Pleco, maker of Chinese dictionaries for smartphones, announces a very sexy OCR add-on for the iPhone. If the final product is anything like the demo, it’ll be blazingly fast. Watch the dictionary entries come up as the phone is panned across the text.

You can also change the OCR window size, varying it from one character to a whole phrase.

It takes advantage of the better video camera on the iPhone 4; I’ll be interested to see how it performs on my 3GS (with iOS4).

Of course they’re demoing the app recognizing text in a book. But you know I’m going to take it out into the world and try it on signage. Product labels at 99 Ranch grocery. Marquees in Oakland Chinatown. Specials on the Chinese-only menu! Pleco Camera Recognizer doesn’t look like an augmented reality app, but it is. Or it could be.

August 12, 2010 at 7:46 pm 4 comments

Twitter Updates

Recent Posts


%d bloggers like this: