Mood Map -- a visualisation of the world's mood
Data visualisation is a strange art. You almost have to approach it like a science.
When you begin, there's a huge mountain of data lying there. Opaque, impenetrable. In order to make sense of it you have to chance a guess; form a hypothesis. Once you have that hypothesis you have somewhere to begin; a way to start analysing your data. And it isn't till you've finished analysing it that you know whether there's anything worth visualising. It's one big fishing trip.
And so it was with Mood Map.
My hypothesis: Key events that occur in the real world would be reflected in people's communications (Twitter). When there was an earthquake there would be global empathy. When there was a world changing announcement, a global rejoicing.
My method: Every minute, sample the public timeline of Twitter for tweets with positive or negative emoticons. It's not a particularly foolproof way of measuring mood, but hey, I'm not a statistician or a text analysis specialist. Once I've got the tweets, geocode them and place them on a map, clustered according to volume and coloured according mood.
My conclusion: After gathering six months of data, monitoring world events, and analysing it all through my custom visualisation engine I've not discovered much. (Or maybe that means I have?) There's no empathy, no rejoicing. Everyone's pretty much wrapped up in themselves. (At least on Twitter <sarcasm>News Flash!</sarcasm>)
There's definitely no patterns I can discern on a global scale. You can see Mood Map at the time of the 2010 Haiti earthquake or the 2010 Chile earthquake. In both cases there's no discernible dent in global happiness, however it is possible to notice localised mood effects. For example there's a noticeable red dot in Poland when the Polish president died in a plane crash, and a smattering of unhappy Europeans while Eyjafjallajokull was erupting. But on the whole, the World stays at a steady 85 - 90% happiness day in day out.
I'll keep the data churning over for a while longer and see if any other patterns emerge, but for the moment it seems like it's best as a tool for spotting local (country-wide) mood fluctuations. If you take a look through the archives and spot something interesting, let me know!