A group of researchers from Northeastern University and Harvard University have analyzed the sentiment of 300 million tweets over a period of 3 years. This is enough data to give us a pretty accurate picture of how Americans are feeling as they tweet throughout the day and week.
This tweet data was then compared to the Affective Norms for English Words (ANEW), which “provides a set of normative emotional ratings for a large number of words in the English language.” They also mashed up that data with information from the U.S. Census Bureau, the Google Maps API and more. What they ended up with is a visualization showing our nation’s moods over time. Only according to Twitter, of course.
The data is pretty interesting. Some of it is as we expect. Our moods take a dive around Thursday and we are much happier on the weekends than during the weekdays. Looks like most of us aren’t too happy to be at work and that whole hump day thing seems to have an adverse affect on how we feel. We are also happier early in the day and later in the evening, presumably because of that whole work thing.
Where things get more interesting is in the geographical representation of the data. The researchers used density-preserving cartograms to display the data. These are maps that take into account the volume of tweets from an area when drawing the map. What you end up with is a map that represents each area’s size by the number of tweets coming out of it, which makes it easy to see which areas are tweeting the most at any given time.
From the geographical data, we see that people on the West Coast are generally much happier than people on the East Coast three hours ahead. East and West coasters follow the same general pattern, but tweeters on the West coast just didn’t get as emo as people on the East. One exception seems to be Florida. Watching the video, you can see that it doesn’t dip as low in mood as the rest of the East Coast, keeping pace with West Coast for the most part.
By taking a look at some basic rankings from the Census Bureau we can take things a step further. States on the East Coast have the largest populations of black people. These include New York, Maryland, North Carolina, and Virginia. Does this mean that black people have bad attitudes, or is this as flaw in the sentiment analysis? Sentiment analysis is not an exact science. It doesn’t understand things like sarcasm or slang. I doubt Affective Norms for English Words is able to pick up on these things either. Given the large population of black people on Twitter, this may have had some effect on these results.
via Time-Lapse Twitter Visualization Shows America’s Moods