Sentiment Sleuths

Want a quick estimate of what Twitter users are tweeting – and feeling – right now, on virtually any topic?

NC State has an online tool for that.

Computer scientist Christopher Healey and his students capture online sentiment through Twitter Viz, a Web tool that graphically displays the emotions expressed in tweets and maps them based on the pleasantness or unpleasantness of the emotion and the emotion’s intensity. Virtually any topic – from cat videos to campaign promises – can be assessed.

Twitter Viz And Elections

During this election season, Healey has focused on mapping Twitter sentiment toward the campaigns of President Barack Obama and Gov. Mitt Romney. The Twitter Viz tool grabs the most recent tweets that include the terms “Obama” or “Romney” and searches for any of 1,100 keywords that provide clues to the sentiment of the tweets. It then maps those tweets on a scatterplot graph, providing a quick sketch of real-time Twitter sentiment.

Hovering over individual points on the graph allows users to view actual tweets in real time. The 500 most current tweets appear on the graph and then cycle off, changing the map every few minutes or seconds, depending on how fast the fingers are flying.

Healey’s program also captures tweet frequency, so users can see exactly when tweets on a given topic increased or decreased.

Twitter Viz And The Debates

Healey uses his tool to examine sentiment during presidential and vice-presidential debates, and to show how debate sentiment changed when compared with the sentiment before or after debates.

Obama and Romney tweets were relatively even during the run-up to the first debate, Healey said. But Romney’s tweets were more numerous than Obama’s during Romney’s reported “victory” in that debate.

During the vice-presidential debate, more people tweeted about Vice President Biden than Rep. Ryan.

Interestingly, Healey says that about two-thirds of the presidential and vice-presidential debate tweets were pleasant or positive.

The tool can even drill down to certain topics mentioned during the debates. If there’s a question about tax policy, for instance, a user can type “taxes” into a search bar and see only tweets that mention Obama and/or Romney and taxes.

It’s Not Perfect …

The tool is not perfect, Healey admits. Poor grammar and sarcasm can skew the tool’s ability to accurately gauge sentiment. And if tweets about Romney and Obama fail to include one of the 1,100 keywords that capture emotion and sentiment – keywords that are part of the Affective Norms for English Words dictionary – the program ignores them.

“I have to stress that although the tool provides good estimates of what the Twitter sentiment is at the time,” Healey says, “during high-volume events like the debates, Twitter only gives us a representative sample of all the tweets on our debate topics. However, the tool does provide an immediate and visceral response which, if you’re interested in sentiment, can be valuable.”

While Healey created customized tools for the presidential and vice-presidential debates – complete with red diamonds for Republican tweets, blue circles for Democratic tweets and green squares for tweets that mention both candidates – the tool can be used for almost any topic, from strife in Syria to Super Bowl games to Lindsay Lohan’s latest exploits.

“Twitter Viz brings to life the types of things you can learn from computer science,” Healey says. “And getting the general public interested in what we study – and why – is an important component of what we do as academics and computer scientists.”