A Blog recording the life a Web Scientist
Digital Futures 2012 – Twitter Communication Analysis
Over the last 3 days, I’ve been working on collecting and analysing the Twitter communications feed for the Digital Futures 2012 Conference, which used the #DE2012 hashtag.
The following analysis is a summaries overview of the work that I’ve been involved in at Southampton University, in a joint project with myself and Edelman ltd. The project, which aims to investigate novel ways of identifying different user roles within online social media platforms such as Twitter has been in development for the past year, during which a number of academic publications, workshops and demos have been achieved. As part of the project’s outcomes, a tool has been developed to explore the dynamic communications of Twitter communications, helping identify how specific individuals are critical to the spread of information. However, one of the biggest problems with visualising and exploring the online communications within Twitter is the sheer volume of information, even within a conference such as Digital Futures 2012. As Figure 1 shows, an unclassified retweet network becomes very cluttered, and it becomes increasingly difficult to identify different user roles as the volume of tweets increase. The aim of the classification model (Described Here: http://eprints.soton.ac.uk/272986/) is to overcome this, making it much clearer to draw out potentially valuable users. The application of the model can be seen in Figure 2.
What follows are the dynamic growth of the networks over the 3 different days of the Digital Futures 2012 conference.
Day 1. Dynamic view of unclassified Retweet Network #DE2012
Day 2. Dynamic view of classified Retweet Network #DE2012
Day 3. Dynamic view of classified Retweet Network #DE2012
As Day 1 shows, viewing the growth of communications in terms of a unclassified network, and ultimately identifying specific users becomes difficult; there is just too much information to comprehend, with multiple network hubs of communication, which seems to indicate a network of disconnected conversations. However, as the videos of Day 2 and Day 3 show, using the classification model, the communications – and individual users – become far easier to comprehend. The Red users, which are those that have been retweeted a certain number times (Day 2 is a minimum of 10, Day 3 is a minimum of 20), are potentially valuable sources of information. However, even more interesting (and potentially useful) are the yellow users who are curators of these highly retweeted users; nor are they a type of user that draws together multiple sources of (potentially) valuable information, but they are also the first in the network to do so. In effect, the yellow users are to some extent responsible for widening the network of communication.
I’ve also included an overview of the timeline of Tweets during the conference and a measure of the top tweeting and retweeted users. What’s interesting is that examining the difference in the highly retweeted users between Day 2 and Day 3 (see videos), nicoleebeale and cyberdoyle change roles, yet the number of curators (yellow users) remain constant.
Hopefully in the next few weeks I’ll be blogging about some new methods that I’ve been working on for performing analysis on Twitter communications, so check back soon!