Monday, July 19, 2010
Following on from "Google flu", a project initiated by employees at Google in 2008, which aimed to monitor flu outbreaks through search, now we have "Twitter flu". Twitter and Facebook status updates are potentially more effective as predictive tools because people enter more words than they would typically do for a search query, so you can get a better idea of context.
Scientists at the Intelligent Systems Lab at Bristol University in the UK collected a database of over 50 million geo located tweets and, by identifying which words were associated with higher incidences of flu, used the data to generate a predictive model.
Any sort of viral effect could be predicted in this way - not just flu but contagious ideas, films or anything that generates chatter in the social space.