(Hey folks – there’s a better formatted version of this post, which includes the flash movies on a Harvard server – feel free to check that out and see the pretty version. I wasn’t sure how my currently ailing blog would deal with embedded flash files…)
Intelliseek’s Matthew Hurst, friend and collaborator, emailed me a couple of days ago with an intriguing idea – if we animated some of the media attention data I routinely collect from Matthew’s site, Blogpulse, would we see a clear pattern of bloggers writing about the tsunami and aftereffects? For the past year or so, I’ve been collecting data from Blogpulse (and about a dozen other sites) by looking for mentions of country names (”France”, “Guinea-Bissau”) in news stories or blog posts and mapping the results directly, and in relation to a number of statistical models.
So what do we get from putting this data into animations? Well, the results are more complicated than one might hope they would be…
Here’s an animation of data from Blogpulse for roughly four weeks bracketing the tsunami. Shades of red on the map represent areas that have high attention within the blogosphere – nations in the deepest shade of red are responsible for 3.2% or more of all the mentions of a nation my scripts are finding in the blogosphere. Nations in blue are receiving very little attention, fractions of a percent.
We’d expect to see the area around the tsunami light up around the 26th of December. And we do – sort of. Because the data we’re looking at here reflects a two week period (i.e., everything posted 2 weeks before the date on the map), the effect is somewhat delayed, and somewhat muted, though certainly visible.
Here’s Google News data for a similar period. The same problems exist with the animation – we’re watching a phenomenon that’s happening moment by moment in two week slices. Again, though, you can see the attention cluster in Southeast Asia a few days after the tsunami.
I also grab data from Blogpulse on a day by day basis, allowing me to look at who’s mentioned what country on a given day. The data is tricky to use because it’s so variable – it’s very hard to draw conclusions about whether a nation is high attention or not based on a single day’s data – but it gives an animation that helps us see the tsunami’s attention effect a bit more clearly. (Sorry there are no dates on this map or the next few – I autogenerate the map frames then manually add the dates and sliders, and haven’t yet added the niceties to the next few animations… I promise to fix this ASAP)
One strategy for seeing the impact of the event over time is to calculate the differences between results from one day to the next. For instance, while there might be 30 stories on Google News on 12/25 about India, there are likely to be 300 on 12/26 because of the tsunami’s impact – if we track that change, by making positive changes red and negative changes blue, we can see how the tsunami’s impact affects attention in the media and blogospheres.
Again, we’ve got a slightly confusing effect – the difference between two day’s sets shows us the effects of the new day… but also of the day two weeks ago we dropped from the data set.
And here’s the blogpulse daily data, tracking intraday changes.
It’s going to take me a good bit more of staring at these animations before I can make some definitive statements about how attention in one area affects attention elsewhere. You’re welcome to stare with me – post something in the comments here if you’ve got some observations to share.
(Big thanks to Matthew Hurst and his team at Intelliseek, and to Nate Kurz, for their help with ideas and tools…)