Follow Datanami:
February 27, 2012

The Science of Social Sentiment

Robert Gelber

Last night while Hollywood’s rich and famous attended this year’s Academy Awards Show, a new analytics tool was hard at work tracking reactions of the Twitterverse.

A joint collaboration from IBM, the LA Times and the USC Annenberg Innovation Lab, produced the Senti-meter a high profile example of the power behind sentiment analysis.

The Senti-meter has a fairly simple graphical interface, broken down into sections of actors, actresses and movies, all with their own “slice” on the graph.

Each slice contains a circle whose size is determined by how many mentions the movie or actor has over time, and a vertical axis shows if the tweets were positive, negative or neutral.

The meter pulls data from Twitter daily and runs language recognition algorithms on the tweets to measure how positive or negative public opinion is regarding an actor or movie. The tool is seen as an example of the role analytics can play in understanding public sentiment through social media.

This is not the first time the public analytics tool has been used to track highly rated events. During the 2011 World Series, the Senti-meter found more than 56,000 tweets showing a higher sentiment for Texas Rangers fans as compared to only 11,500 for the St. Louis Cardinals. It was also used to track the fan favorite QB of Super Bowl XLVI (Eli Manning had a higher rating than Tom Brady off the gridiron as well).

Google and Yahoo played their own Oscar game, attempting to predict this year’s best picture. Pulling search data from users in New York, and checking back with search results from last year, Google said the best picture would either be The Artist, Incredibly Loud and Extremely Close, Midnight in Paris, or War Horse. Yahoo on the other hand, decided to go all out and correctly predicted The Artist as best picture, giving it an 89.7 percent probability of winning. Yahoo did not however, poll their own data to get this result, instead they pulled data from sites like Belfair and Intrade using a “follow the money” strategy that ultimately worked for their prediction.  

While all of these predictions and sentiment tools display new facets of analytics, the technology behind them could influence business and public policy. Instead of setting up a focus group and getting potentially filtered results, a business could track sentiment surrounding a new product announcement or an ad campaign from a competitor using information from social networks.

Datanami