Error: Twitter did not respond. Please wait a few minutes and refresh this page.
A Blog recording the life a Web Scientist
Twitter has become an integral part/tool/distraction during academic conferences, supplying an endless stream of communication between participants, feedback and praise (sometimes criticism), and even as a ‘news ticker’ for those not able to take part and attend the conference.
As with any large streams of information (conferences being no exception), it is often hard to keep track of what is being said, who is producing the important and valuable content, and what is just back chatter (i.e. what’s for lunch or where the ‘afterparty’ is). Finding the important information is always a challenge as more often than not, especially when trying to obtain a concise but comprehensive overview of a conference.
As part of my research which i have been demoing (Identifying Communicator Roles within Twitter), I’ve been working on ways to help identify different user roles within topical Twitter conversations, helping interested parties to who may be the users to follow (or target!). The model (which I won’t go into detail here, come and talk for more information) is based upon filtering users based on the dynamic network retweet network structure that occurs as conversations occur between users (bound by a specific hashtag). Examining the timeline of these tweets the model can be applied and different user roles start to become identifiable.
After receiving a good amount of interest today, I thought a good way of demonstrating my work would be to show the #ICTD2012 twitter retweet conversation network graph, both unclassified and then classified. The unclassified retweet network, shown in figure 1 (forget the colours and size, all nodes are the same) is all the retweets captured during the ICTD2012 conference during day 1 and 2, as it can be seen, it is very messy and identifying potentially important users becomes a difficult task. In comparison to this, Figure 2 shows the same dataset with the model applied to it, and immediately it is much clearer, and certain users begin to become identifiable. This time, the red nodes (@meowtree and @RitseOnline) are those users that are being highly retweeted (which in this graph the minimum retweets needed to be a red node is 40). More importantly, the orange nodes (@Anandstweets, @ekisesta, @Katrinskaya, @katypearce, @virbrussa) are the ones that are actually connectors between these highly retweeted users, potentially users that might be a good source to follow for an aggregated feed of news! What is really interesting though is how this will change over the next few days, will their roles stay the same, will more red and orange nodes start to appear? Something that only time will tell!
This is obviously a very brief overview of the concepts that underpins the classification model, which is still in its very early stage of development, but the applications of this for the ICTD community could be beneficial in the future.
Stay tuned for another look at the #ICTD2012 Twitter conversation towards the end of the conversation; let’s see how the network changes in the next two days!