Following on from the data posted on the USC Public Diplomacy blog, a few people asked why not include #IranElection – the answer was twofold, first it had so much more traffic it would have dwarfed the others, making distinctions harder to identify. Second, the level of traffic limited the extent to which a user could keep up with the flow of information from #IranElection. at its peak the #tag was running at over 22,500 tweets an hour and nearly 100,000 tweets in a day. There is little anyone can get from reading 375 tweets a minute, forcing users to rely on filters such as the more specific tags, or choosing only to follow certain users; narrowing the field of view but having a chance to understand what is being produced.
Continuing on from that post, and with a nod of thanks @jobrieniii for the work on Twapperkeeper, I have looked at the data from #IranElection for 10 weeks following the election.
The following images chart the usage of #IranElection and shows 1.4 million tweets spread out over that time.
I wouldn’t quite say I’ve got all of the bugs out of the data,so consider these ‘draft’, but they give a good idea of how usage has varied with time.
Below is the comparison of #Neda and #GR88 usage over a slightly later time period. A further observation form the data is that the top two contributors to #neda have used the tag over 4,000 times each.
The image below, backing up the initial observation in the USC post that users of #HelpIranElection were isolated from other discussions, demonstrates the longevity – or should that be brevity – of usage this #tag had amongst users.
While still vaguely active in October, Tweets a day had barely been in double figures for quite some time. Actual interaction through this tag was also much lower; 78% of tweets carrying #HelpIranElection were identical.
The average tweets per user suggests a similar story. For example, in the case of #FreeIran the average tweets per user was 15.4, #Sohrab it was 11.99 while #HelpIranElection was used on average 1.09 times. This is the difference between ‘click to show your support’ of #HelpIranElection and genuine interaction.
Will update further soon.
Interesting but raw data. I would say that in its current state it is not very informative, however. A few suggestions:
– you are quantizing your data on a day (70 buckets) but in early days information was unrolling hour-to-hour. Might be more useful to have hourly buckets, perhaps overlaid with the daily graph for smoothing purposes.
– The graphs would be immensely more useful if you labelled the X axis with important real-world events. This would tell the reader what all the secondary spikes are in the first graph.
– how about a secondary plotline showing the number of active unique ‘tweeters’? What is the %age of retweets? How may tweeters only retweet?
More as I think of them. You are just scratching the surface.
Thanks for reading and your comments, as you suggest there’s a lot to look at, which is why we’ve been following the data. As the note from EA suggests it is interim – in the post I highlight it’s draft. I expect by the time we get done there’ll be quite a few different looks at the data, some of which will be as you suggest, there may even be a look at time-series mapping, visualising links growing and comparing that to tweet volume.