Using Twitter to analyze mood in Canadian cities

Posted on Friday, December 19, 2014

Are some cities happier than others? Do single events dictate the mood of entire metropolitan areas? By analyzing tweets and the emotions they convey, researchers at the University of Ottawa hope to increase the knowledge smart cities have about their citizens' quality of life.

Putting theory to practice, researchers at the University of Ottawa compiled data from six cities—Ottawa, Montreal, Vancouver, Edmonton, Toronto and Halifax. The data comprises 132,181 tweets from December 4 to December 10, 2014.

A summary of the most interesting findings:

  • When the Montreal Canadiens won over the Vancouver Canucks earlier this month, “happy” tweets by people in Montreal were 69% higher relative to those by their western counterparts.
  • Halifax is Canada's party city, with happy tweets on Friday evening and the weekend 17% higher relative to those posted by Vancouver residents during the same time.
  • On average, Edmonton is the “least happy” city of the six, with most cities surpassing the proportion of happy tweets in Edmonton by 5% to 12%.
  • Positive tweets were 13.8% higher in the morning in Vancouver and Montreal relative to Edmonton, while Ottawa (7.8% higher relative to Edmonton) and Toronto (8.6% higher relative to Edmonton) take over during the day.

“Social networks have been shown to provide more reliable information than traditional and time-consuming data collection methods like questionnaires. Think of it as implicit feedback,” says Benjamin Guthier. Guthier is a researcher at the University of Ottawa's Multimedia Communications Research Laboratory (MCRLab), led by Abdulmotaleb El Saddik, director of the lab and a professor at the Faculty of Engineering.

The researchers used a four-dimensional vector analysis involving the four elements of pleasantness, arousal, dominance and unpredictability. Researchers also used emoticons and hashtags when collecting messages. In total, they compiled and categorized 47 different “emotion-word hashtags” (e.g., #angry, #happy, #disgusted) as well as their relationship to the other words in the tweet.

“Twitter is rich with emotional content. It has been used successfully in studies about disasters and emergencies, to monitor public health as well as in exploring events and, more recently, in sentiment and emotion analysis….Having access to this kind of information would be particularly helpful for important policy and other decisions,” states professor El Saddik.

Media inquiries

Kina Leclair
Media Relations Officer
Office: 613-562-5800 (2529)
Cell.: 613-762-2908

Back to top