Urban Narrative - Tracking Movement via Twitter

We have been logging Twitter activities in the London area (M25) over an earlier weekend in January with some code Steven Gray has put together. The idea was to log the location based traffic and see what the mapping of it would bring. There are a number of twitter mapping projects out there already, for example twittermap.tv from where the timeLapse of the weekend activity was captured HERE, or the first big mapping project twittervision.com. However, we wanted to focus on a local region, a city, to see what the traffic is and how the location might play a role. The traffic visualisation page tweeTOMeter is part of this interest.


Image by single-blogs / Screenshot of Twitter data visualised in GeoTime’s space-time aquarium.

One could think of this investigation as following the urban story quite literally, while following the tweets of citizens. However it is quite tricky to make sense of it all. The dataset for the weekend, which covers Friday evening to Monday morning contains some 300’000 tweets. Not all of them are properly geo referenced. Only 1’700 have actual Lat/Long information in the geo tag field. Furthermore some 60’000 have Lat/Long details in their profile tag field and the ret only has a generic profile location, such as London. This probably is because of the relatively new geo support of the Twitter API. Most users still seem to have little interest to include their actual location, as well as a lot of the applications do not yet properly support the format. Interesting seems to be the network. Whom are tweets directed at? It seems to be quite a high average of direct tweets, almost 3 per message. Also who will actually read it, how many followers are there in average?
Working with the real geo referenced tweets, surprisingly they contain quite a bit of movement.
For a quick look at the data it has been visualised in GeoTime. The representation in the time-space aquarium makes the diagonal lines, that suggest movement, very distinguishable from the vertical stationary lines. While looking at the replay in the 2D view the weekend really comes to life and London gets busy.



Similar visualisation, with snippets and names, but without the river Thames, can be fund HERE.
GeoTime here really offers a powerful and very quick way of visualising the data in space and time and offers a whole pallet of different visualisation types, each including a set of tools for analysis and manipulation. Import comes either via ARCGIS or even quicker excel.
The main problem really is the quality of the graphics, the design of the result. Here the user has hardly any choice or possibilities to manipulate anything from colour palette to line style or font. This is a bit annoying especially because the tool is kind of an end of the line analysis tool, after you have prepared the data elsewhere.
The second quick one goes into Google Earth obviously. Here the data again comes from a simple excel spread sheet with a VB macro to write the KML file. This literally takes 5 seconds to do and you have a KML file, including time tags in Google Earth.
This one only plays the locations though, also in a time window of some six hours.