Identifying the sentiment styles of YouTube’s vloggers Bennett Kleinberg Maximilian Mozes Isabelle van der Vegt Department of Psychology Department of Department of Security and University of Amsterdam Informatics Crime Science Technical University University College London Department of Security of Munich
[email protected] and Crime Science
[email protected] University College London
[email protected] Abstract examine natural language in this young field of communication. Nevertheless, little attention has Vlogs provide a rich public source of data thus far been paid to investigating the language in a novel setting. This paper examined the used in YouTube vlogs. continuous sentiment styles employed in Much of the literature concerning Youtube vlogs 27,333 vlogs using a dynamic intra-textual approach to sentiment analysis. Using focuses on the visual modality (Aran et al., 2014) unsupervised clustering, we identified or meta-indicators such as views and subscriber seven distinct continuous sentiment counts (Borghol et al., 2012). With the current trajectories characterized by fluctuations of study, we aim to address this gap in the literature sentiment throughout a vlog’s narrative by automatically analyzing the linguistic styles time. We provide a taxonomy of these used by YouTube’s vloggers. Building on a novel seven continuous sentiment styles and approach to examining the continuous sentiment found that vlogs whose sentiment builds up structure, we seek to shed light on the different towards a positive ending are the most temporal trajectories used by vloggers, and, by prevalent in our sample. Gender was doing so, we expect to gain a deeper understanding associated with preferences for different continuous sentiment trajectories.