Exploring Characteristics and Behaviors of Live Streamers

Live streaming platforms such as Twitch, YouTube Live, and Periscope have become some of the most popular synchronous social networking services. With the increasing prevalence of live streaming services, intensive competition has emerged among streamers on live streaming platforms. In the meantime, as a superstar market, live streaming suffers from competitive bias toward popular streamers, leading to a potential loss of small streamers and an over-priced sponsorship for advertising. As such, there is an urgent need for streamers and platforms to address this challenge. In this dissertation, I examine this challenge through the lens of the heterogeneous and dynamic competition of live streamers. The underlying approaches are designed to understand how streamers' characteristics and competitive behavior affect the popularity and viewership of streaming channels. Relying on measures mined from streaming speech and video, I first empirically examine the key factors that determine streamers' ability to attract large audiences, namely their personality, professionalism, and streaming affordance. Based on the unique dataset of channel information and the viewer lists drawn from the live streaming platform Twitch.tv, I then investigate how competitors' exogenous entry affects the popularity of streaming channels. In addition, I propose a context-aware deep learning framework to extract embedded live streaming subframes and empirically examine the role of video quality and streamers’ emotional expression in real-time viewership. This research constitutes contributions toward understanding the nature and dynamics of live streaming. The result provides comprehensive metrics and useful insights to evaluate streamers' popularity in the competition environment, which in turn generates actionable strategies for social media influencer investors, live streaming users, and platform operators.