Every tweet, no matter how trivial, reveals the writer’s mood, through word choice–It can be as obvious as “happy” or as subtle as “diamond.”
Now imagine if you could take that knowledge, scale it up to the entire Twitter-verse, and use it to gauge the entire country’s mood? That’s exactly what Alan Mislove, a computer scientist, at Northeastern University, did, using 300 million tweets in real time. Check it out in action, over time–you can actually see moods rising at the end of the work day, and that same pattern gets repeated across time zones:
Though it seems impossibly hard, the methodology was actually quite straightforward. Mislove and his team started with a scale, previously devised by psychologists via polling, which assigns mood scores to 600 English words.
Mislove then used that scale to analyze 300 million tweets that originated in America, from 2006-2009. And he was then able to create a collective mood score for each state. (The reason the map above is so lumpy is that the area of each state corresponds to the number of tweets originating there.)
The only real barrier to tech like this from becoming leaps more subtle and profound is the word-scoring method. With a better, finer grained way to parse words and messages, you could easily imagine a tool that reveals emotional responses to events–and even uses that data to make market predictions. (And Twitter buzz actually turns out to be an uncanny predictor for how well a movie will do at the box office.)