And its raining
Every day, two hundred million people post minutiae of their lives to Twitter. Mathematicians at Bristol University have developed ways to turn this mass of mundane messages into an event tracking tool.
Two hundred million participants on Twitter does not represent a good scientifically chosen sample, but the geographic spread and timely nature of this data means that researchers Professor Nello Cristianini and Vasileios Lampos have been able to develop algorithms to monitor twitter to detect events. Their first work, based on samples of 200,000 tweets from 49 urban centres in the UK, enabled them to track the presence of colds and flu, and to compare those real-time rates with statistics compiled by the NHS, with surprisingly good correlations. This opens the possibility that Twitter data could be used as an early indicator of outbreaks of flu epidemics.
Their latest work ties in with the British fascination for the weather, and the extraordinary number of people who need to tweet to the world that its started raining. Again, this work can be cross-checked against weather stations throughout the UK and shows that Twitter postings can be used to detect and measure events in real time.
Now this raises the question of what else could be monitored this way, and how long will it be before this becomes a standard tool in the social research toolbox. For instance, when we have elections, will Tweets provide clues to voting intentions? For people in the distribution business, can Twitter tell them where supply is failing to meet demand? Marketing gurus have long said that organisations should monitor social media, but these researchers have shown how such generalisation can actually be achieved.
29th November 2011