June 06, 2012

#LexingtonPoliceScanner Twitter posts

When the University of Kentucky beat Louisville in the Final Four of the NCAA Men's basketball tournament there was literally rioting in the streets.  Two days later (April 2, 2012) there was a repeat of events.  We are happy to report (thanks for asking) that the entire Floating Sheep team was unharmed by these events and were either calmly sipping a glass of Chardonnay or running wild in the streets on-site documenting the events as they unfolded.

Of particular interest to us, however, was the use of the hash tag #lexingtonpolicescanner within Twitter to record the ongoing events those nights.  We missed the data on first riot but download the Tweets from the second one (about 11,000 over the course of four hours during the evening of April 2nd and early morning of April 3rd).

We were curious about what the data within the tweets could tell us so we first took a look at what words should up in the 140 characters of all the tweets.  We removed the hash tag (#lexingtonpolicescanner) as well as "RT" which stands for retweet and made a wordle cloud out of it.  A fairly unsophisticated analysis but it brings up some intriguing results of what was capturing the interest of the portion of the Twitterverse interested in Lexington that night.


Of course it also showed some obvious other words to filter such as Lexington, Kentucky and Twitter. When these common terms were removed, the Wordle cloud zooms more into focus with clear references to "fire" and "couch" and "car" which the combination provide some of the big events of the evening.  Why people celebrate by burning furniture remains one of the mysteries of the universe. Other interesting terms include "fireworks", "nude", "street", "fans", "shooting" and "entertainment". We'll leave the rest for you to interpret.


More intriguing (and a big part of this blog) is the geography of these tweets. Using user specified location (available for about 50 percent of all user profiles) we map the location of tweets referencing #lexingtonpolicescanner.  See the map below or try out the interactive version (see more information about this at the end).

Distribution of all Tweets with #lexingtonpolicescanner from April 2 to 4, 2012

Interestingly the map showed a relatively similar pattern of interest in Kentucky basketball as did our analysis of NCAA basketball nations from two years ago. The map below shows points in the U.S. where Google Map searches produced the most hits for an array of basketball teams including Kentucky (see the original post for the methodology).  While not a perfect match, the comparison of these two user-generated representations of Kentucky Basketball show a concentration centered within the Commonwealth with a dispersion that dissipates according to a fairly standard distance decay function.  Very few points further than a few hundred miles beyond Kentucky's borders show up.

  NCAA Basketball Nations
Finally, we (or more correctly Ate) set up an interactive interface with the data and it is well worth checking out.  You can slide the time bar and see how this database of tweets changes over time and space.  There is an initial rather geographically wide surge of interest as the actual riots play out but by the next morning the spatial extent of tweets to mostly just within Kentucky and Lexington.  As such it is a nice illustration of the temporal-spatial dynamics of a news event.

Note, this is a experimental/working version so it is still rather rough around the edges. It makes use of the versatile D3 visualization library. Since it incorporates all kind of HTML5 goodies, it works best in Chrome or Safari.

No comments:

Post a Comment

Note: only a member of this blog may post a comment.