Many Wikipedia articles are tagged with geographic coordinates. Many have references to historic events. Cross referencing these two subsets and plotting them year on year adds up to a dynamic visualization of Wikipedia's view of world history.
The 'spotlight' is an overlay on the video that tries to keep about 90% of the datapoints within the bright area. It takes a moving average of all the latitudes and longitudes over the past 50 or so years and centres on the mean coordinate. I love the way it opens up, first resembling medieval maps of "The World" which included only Europe and some of Asia, then encompassing "The New World" and finally resembling a modern map.
This is based on the thing that me and Tom Martin built at Matt Patterson's History Hackday. To make it, I built a python SAX Parser that sliced and diced an xml dump of all wikipedia articles (30Gb) and pulled out 424,000 articles with coordinates and 35,000 references to events. We managed to pair up 14,238 events with locations, and Tom wrote some Java to fiddle about with the coordinates and output frames. I've hacked around some more to add animation, because, you know, why not?
The code is here.
I've uploaded the data into Google Fusion tables in case anyone else wants to have a look.
Here's every wikipedia page tagged with a latitude and longitude, a total of 424,171 articles.
You can get a great idea of the static distribution of coords in this heat map plot.
Here are the 14,238 geo-tagged events dragged out of Wikipedia. Many of the events are attributed to a particular year or month, but the day is not specified. I defaulted them to 0 to indicate this. Hopefully it won't matter too much from the macro-perspective at which this data should be viewed.
And here are 5765 invalid coordinates that cropped up. The majority of these are actually invalid, although some valid ones slipped through the net. If I come back to this project, I'll polish up the parser to handle the final few. In the meantime, anyone looking to improve Wikipedia's data could have a go at correcting the bad data referenced here.