NCSU program samples Twitter sentiment during debates
Posted October 4, 2012 4:08 p.m. EDT
Updated October 4, 2012 4:35 p.m. EDT
Raleigh, N.C. — Twitter users let fly with some 10.4 million tweets during the presidential debate Wednesday night, and, contrary to social media stereotypes, they weren't all nasty reviews of the two candidates.
That's according to a program developed by North Carolina State University computer scientists.
Associate Professor Christopher Healey deployed his Tweet Viz application on WRAL's servers Wednesday night to measure Twitter sentiment during the presidential and North Carolina gubernatorial debates.
"It surprised us a little bit that it was trending positive," Healey said.
The program uses a "sentiment dictionary" to correlate words used in tweets with emotions. The program estimates how intense someone is feeling as well as whether their language is more or less pleasant.
Wednesday night served as a test run for the software, which Healey and WRAL hope to make available during upcoming debates.
"Twitter gives you a really nice way to capture people's emotions at the point and time the debate is unfolding," Healey said. Unlike blogs or even Facebook, Twitter users tend to post about what is happening in front of them, providing a sample of real-time opinions. The effect is not dissimilar from a focus group "dial test," in which a small group of voters is monitored for their reaction to candidates.
Fracking gets Twitter users' attention
Healey's program monitored 2,949 tweets sent regarding the gubernatorial debate between Lt. Gov. Walter Dalton, a Democrat, and Republican Pat McCrory, a former Charlotte mayor.
A review of peak topics discussed during the debate seemed to show natural gas drilling, education funding and a controversial ad by Dalton targeting African-American voters grabbed Twitter users' attention.
"The one that we noticed was fracking," said Healey, using the shorthand name for the method of extracting natural gas through horizontal drilling and hydraulic fracturing.
Dalton said that he would back fracking if he were convinced it could be done safely. McCrory said he was convinced fracking was safe and said the state should open itself to the process.
Twitter users traded thoughts and summaries of those two tactics, but one particular line was often quoted.
"Spoken like someone who represents big oil," Dalton said in response to McCrory's assessment. That was a swipe a McCrory's work at a law firm which represents the American Petroleum Institute.
At first blush, that doesn't sound like a terribly positive thought. But many of the tweets about that line were Dalton backers cheering on their candidate. McCrory backers seem to brush off the line.
That seemed to be a trend across both debates – more users took to Twitter to cheer on their candidate than to slam the other guy.
Throttling the presidential debate
Originally, Healey planned to monitor all tweets relevant to the presidential debate, but Twitter "throttled" the number of tweets available to the program. Still, Tweet Viz was able to randomly sample 350,000 tweets, roughly 3 percent of those that related to the debate.
The more free-wheeling nature of the contest between President Barack Obama, a Democrat, and Republican Gov. Mitt Romney makes it harder to match up reaction to what was going on at the time. However, one interesting trend did jump out.
More Twitter users in the sample, to the tune of 2-to-1, were tweeting about Romney than Obama.
"It seemed more supportive than trashing," Healey said.
That would be consistent with the general impression that Romney did better in the debate than Obama.