A comparative evaluation of techniques for time series visualizations of emotions

The growing availability of social media and other online information sources has increased interest in sentiment analysis to understand the emotional responses of users. Being able to visualize users' emotions could help stakeholders to better understand the underlying trends behind events or stories. Various techniques have been used to generate time series visualizations of emotions; however, there is neither a prevalent method nor any guidelines for the design of visualizations that depict emotions and their evolution over time. We report on a controlled user study that compared four common visualization techniques. User performance and preferences were measured under a formal task taxonomy, using Twitter data about real-world events. The results, although highly task-dependent, show both an overall performance advantage and a higher level of user preference for the line chart, and suggest that the radar chart, despite its popularity in the literature, may not be the best choice to depict such data.