While Twitter users mostly use the microblogging service to shoot back their thoughts on whatever they happen to be eating or marvel at the dadaist poetry of a spam bot, researchers can’t seem to get enough of the service. For them, it’s become an enormous trove of information about what people are thinking or feeling. One U.K. research group has found a strong correlation between major events and moods on Twitter, and suggests that this information could even be used to predict major events — like the riots that rocked the U.K. last summer.
The research, led by Thomas Lansdall-Welfare from the University of Bristol, calculates the mood of a nation by looking for specific key words which the researchers believe are indicative of certain moods. Using a piece of software called WordNet-Affect, the researchers sorted words into four categories — joy, sadness, anger, and fear. Using this methodology, the team looked at some 484 million tweets generated by 9,812,618 users between July 2009 and January 2012.
Not surprisingly, they found a strong correlation between the mood calculated by their model and events at the time. For instance, the team says that Christmas sees a strong spike in the “joy” category, with smaller spikes around New Years and other holidays. The team also said that last April’s royal wedding between Prince William and Kate Middleton also brought a swell of national joy on Twitter.
Most interestingly, the team noticed a strong rise in their “anger” category last summer. This peaked around August 6, 2011, the first of several days of rioting. Though there were numerous public indicators of mounting unrest leading up to the riots, that it was so strongly correlated on Twitter is certainly worthy of note. Not only that, but the researchers also found similar spikes in 2010, at the announcement of heavy budget cuts by the government.
Of course, we’ve seen this kind of research before. Like those studies, this one is limited by the size of the Twitter using population versus the total population, and the subjective nature of the analysis. For instance, a large number of obligatory “Happy New Year” messages could throw off the model. Moreover, the system has no way to deal with sarcasm, and takes all Tweets at face value.
What this study has shown is that interesting, and perhaps useful, data can be pulled from Twitter using relatively simple means. The group has even put a tool online that aims to function like a Twitter-sourced U.K. mood ring. At the time of writing, it shows that the U.K.’s joy is down -24.12, with an ever-so-slight uptick in fear. Does the U.K. need a hug?
- According to Twitter, everyone is getting sadder
- Global mood swings, via Twitter
- Twitter might be a new stock market bellwether