TopicWave: Interactive Visual Analytics of Spatiotemporal Topics Distribution of People‟ s Reactions to Events from Geo-tagged Social Media

Siming Chen, Jie Li, Gennady Andrienko, Sankt Augustin, Nathalia Andrienko

Abstract:Messages in social media, such as Twitter and Facebook, are often posted in reaction to some events happening in the world. People produce text messages with semantic contents (meanings) at different moments in time being at specific locations. Understanding the variety of people‟ s reactions to significant events is important in many areas, such as journalism, social science, business, and politics. Our goal is to propose a way to analyse the relationships of the message contents to the space, time, and people. It is challenging due to the large volume and multifaceted structure of the data. Visual analytics is a science aiming at supporting human analytical reasoning by interactive visualisations combined with capabilities of computational processing (Thomas and Cook, 2005). To enable the spatial, temporal, and semantic analysis of event-prompted reactions of social media users, we propose a visual analytics system TopicWave, which combines computational techniques for topic modelling with interactive visual tools enabling the exploration of the variation of the derived topics across the space, time, and set of users.Recent state of the art in visual analytics regarding analysis of social media data is summarized in (Chen et al. 2017a). Andrienko et al. analyzed social media from spatial-temporal and topic perspectives to investigate users‟ mobility (Andrienko et al, 2013). Chen et al. proposed a map-based visual metaphor to visualize the event evolution (Chen et al., 2017b). Unlike the previous works, we focus on the evolution