As Newt Gingrich took the stage Saturday in South Carolina to the cheers of enthusiastic supporters, tweets about him rose sharply, as did expressions of shock, according to an analysis by San Francisco-based Kanjoya.
“If Barack Obama can get reelected after this disaster, just think how radical he would be in a second term,” Gingrich said.
Gingrich's victory and his words afterward took many people by surprise, Kanjoya reported. Its engineers found tweets such as: "You never know what will happen until it does. A bit surprising. Gingrich Wins South Carolina Primary."
Although some of the "surprise" tracked might be read as either neutral or positive, others on Twitter seemed stunned in a negative way: "Watching Newt Gingrich speak.. Does he really think he has a chance of becoming president of the U.S.???NOT!!!"
There was another element in "surprise" sentiment tracked by Kanjoya: Surprise that his marital history, which includes two divorces and his admission of infidelity, had not damaged his chances in conservative South Carolina.
"Just find it unfathomable that Gingrich won the family values vote," wrote user Justin Barbour.
At 6:30 p.m., when Gingrich began his victory lap in front of backers in a South Carolina ballroom, Kanjoya had analyzed about 183,000 tweets associated with him on Saturday. An hour and a half later, more than 288,000 tweets had been tied to Gingrich by Kanjoya's computer algorithm. By 9:30 p.m., tweets analyzed had risen to more than 310,000.
The company is using an algorithm to assess emotional language that was derived from the Experience Project, which has collected more than 14 million personal stories, with their associated emotions, since launching in 2007.
Kanjoya's analysis Saturday acknowledged limitations in sentiment analysis done by an algorithm. A computer might assign an emotion to a tweet, but it still takes a person to decide what it means.
Moritz Sudhof, a Kanjoya engineer, noted that although expressions of "joy" rose for both Mitt Romney and Gingrich as the polls closed, "we see two very different stories."
Tweets about Romney that were assigned "joy" by the algorithm often also contained the words "tax returns," which Kanjoya believes is an indication of pleasure regarding his loss in the wake of his decision to wait to release his tax returns, Sudhof said.
And consider this, in tweets about Romney: "The top distinctive term is 'lose' -- so it seems like people are pretty happy that he lost," Sudhof said.
Sudhof said Kanjoya found that tweets immediately after Gingrich was declared the winner based on exit polls were dominated by "words like 'victory,' 'leads,' and 'voters' -- much more standard 'we're happy he won' kinds of tweets."
Later, Sudhof compared tweets between 8 and 9 p.m. PST with those of the previous hour to see what was driving the rise in the volume of Gingrich tweets.
"From the joyful tweets, we see a change in the kind of conversation. Two hours ago, the joyful tweets are distinguished by words like 'delegates,' 'winning,' 'statewide,' 'district' -- so, very election-specific words," said Sudhof.
"Now, the joyful tweets are much more celebratory," he said. Distinctive terms found by Kanjoya included 'resurgent,' 'kickin,' and 'beat.' ”
Read more about Kanjoya's algorithm: What sentiment analysis can and cannot tell us about the GOP race