By registering normal user activity on social networks we can now link a change in activity patterns to the arrival of a new buzz topic.
According to three researchers at the University of Tokyo, if you want to rapidly identify topics of interest on social networks, the solution is not to sift through all message traffic for the most frequently-used words. It’s better to observe the behaviour of a focus group: when there’s a deviation from their normal publishing habits, that can be a sign that a new buzz topic is emerging. The researchers have developed a model which focuses on user-interaction. The explanation is quite simple: a discussion topic may arise from a single picture, or from an audio track. In these cases, the traditional detection methods will clearly not work.
Detection based on "normal" behaviour patterns
The software that the Tokyo researchers have developed works on two cycles. Firstly it analyses a number of users from a chosen network (in the tests they used Twitter), and will register their "normal" behaviour pattern (number of posts, how often other people are mentioned, etc). Once this has been done the software continues to analyse the network, but will now be able to identify behaviour seen as abnormal among the user-group. For example, an increase in the pace of tweets, repetition of a specific word, continual messages to an individual or a group in a pattern that deviates from the norm. And if several users start to behave surprisingly in quick succession the software will indicate the emergence of a new buzz topic.
Tests prove conclusive
The test phase has proved conclusive, since using the software has highlighted emerging phenomena at least as quickly as with methods based on spotting key words. And it’s interesting to note that the Tokyo method has been much more effective where key words were difficult to identify (they cite "NASA" as an example). The main conclusion we can draw is that the two methods could well be complementary: the researchers don’t rule out using content search functionality to enhance the effectiveness of their software.