Who Says Social Media Has No Emotion?
Have you ever wondered what someone really meant in the tweets or emails you read on a daily basis? Are they written by someone who means what they say or is angry, apathetic or just devoid of emotion? Or perhaps you are reading too much into the text, which has happened to almost anyone who has been on the receiving end of an unclear electronic response.
The written word can be so hard to discern because of the lack of voice inflection, eye contact and facial expressions. As the newer forms of communication boom in popularity -- texting, emailing and tweeting – the human emotion gets further and further removed from the message. In an effort to analyze and identify those seemingly obtuse emails and tweets, graduate student Siddarth Ramaswamy and undergraduate student Carson Holgate worked with Dr. Christopher Healey (pictured), an associate professor of Computer Science at NC State, to develop a method of identifying the emotion behind social media posts.
Text sentiment analytics or sentiment analysis, measures the emotion in abbreviated messages in such forms of communications as Twitter, an online social network which allows users to communicate short text messages - tweets - of only 140 characters. Healey defines this sentiment as “an attitude, thought, or judgment prompted by feeling.” The goal of the project is to identify basic emotions that are imbedded within text and to measure the degree of those emotions, as well as detect patterns that correspond to the identified emotions.
Research by Ramaswamy and Holgate focused on visualizing the sentiment of tweets posted onTwitter, which averages up to 177 million posts per day. Twitter has been instrumental in changing the communication landscape by enabling users to organize protests across the globe, break critical real-time media stories, and keep the world up-to-date with the latest news.
With the restrictions of just 140 characters per tweet, users must be incredibly succinct in their views, thus leaving much to be discerned by the reader. With that discernment comes the determination of the emotion behind the message. Just what did they mean by what they tweeted?
“Psychological studies show different dimensions of emotional affect from text sentiments. Two important ones are pleasure, and activation or arousal, such as being highly excited by information received in a text sentiment,” Healey said. “A variety of emotional states can be derived from those two main emotions. We can build on that by calculating pleasure and arousal from the tweet text, then based on a scatter plot of these results, we can determine the emotional impact, such as a large number of responses to the left of the scatter plot would identify unpleasant emotional responses, while a large number to the right would indentify positive responses to the tweet.”
The tool is being made available for use by the general public, Healey said. (To see the tool in action, click this "Tweet Viz" link.) The decision to pursue the Twitter text sentiment project is due to the highly powerful influence Twitter has had on the social media landscape.
“Major organizations monitor Twitter 24-hours a day and if a person with a lot of followers tweets that they had a bad experience from Company X, then the typical response is that Company X would immediately call the dissatisfied customer to see how they could fix the problem, in hopes that their actions would result in a favorable tweet and ultimately do damage control to the company’s reputation,” he said. “In that incident, their quick monitoring of the negative emotion could quickly change the negative trending of the incident to a positive one.”
Twitter topics are timely and when people are querying them, they can use the tool to determine what the overall sentiment is of the trending topics of the day. The tool is specific enough to detect different layouts of electronic communication, Healey said.
“When people are talking about specific subjects online, very often they will break down into sub-topics and the tool allows for detection of topic clustering as well,” he said.
The tool could actually be used for detecting emotion in email, Facebook posts, cell phone texts and other electronic communication, Healey said. While it is very intuitive and does a good job of tracking emotions in trending topics, thereare a number of gray areas that it cannot detect.
"One emotion it cannot detect is sarcasm,” Healey said. “It only recognizes a subset of words in the tweet and has no way to determine if what you are trying to get across is meant with a sarcastic bent to it.”
As in life, not all things are crystal clear, even to computers, but Dr. Christopher Healey and his team have paved the way for clearing up some of the mystery behind the world of electronic messages.
For more information on Dr. Healey’s work, visit his website at http://www.csc.ncsu.edu/faculty/healey/tweet_viz/.
Return To News Homepage