Before I went to bed, I wanted one last Australian related election post
. This post draws on the data and methodology from <a href=[edited] other posts</a> and <a href=[edited]
At around 6pm, I did four additional keyword searches on <a href=[edited] beyond the ones in my most recent post. These were: #ausvotes, #donttrustabbott, #nocleanfeeds, #electionwire. I added them to my complete list of Tweets, then filtered the tweets based on username, location and tweet to remove duplicate tweets. This brought the total of 60,0084 Tweets down to 54,962. I then added three columns: City, State, Country. For the next few hours, I endeavored to fill in as many city fields as I could. The focus was on identifying Australian cities. Thus, for countries not Australia, I just listed the country or unknown. This was so I knew to ignore those. At 8pm, I stopped completing the fields. I just don't have time to label everything...
Of those 54,962 tweets, some form of identification was completed for 34,213 tweets or 62% of all tweets. (Not a bad sample size from the whole.) Of these tweets, 19,165 came from Australia. Of these, 15,629 have Australian cities identified for the location of the Tweet. That I can play with.
The next step is to identify noise of Liberal, Labor and the Greens by city. The following CONTAINS filters were created on Excel to find tweets to get location data: Labor OR Gillard, Abbott OR Liberal, Brown or Greens. There were 1207, 977 and 302 tweet locations respectively. After this, the total number of Tweets per city for those terms was counted. There were 61 cities with Labor tweets, 62 cities with Liberal tweets, 38 cities with Green tweets.
Created August 20, 2010 at 10:29 AM
Updated September 28, 2010 at 08:20 AM