Bridging Human-Centered Social Media Content Across Web Domains

Social media is at the forefront of all technologies that have had a disruptive impact on existing infrastructures. Being predominantly human-centered, social media has provided an innovative paradigm to address various cultural (e.g., freedom of speech), sociological (e.g., community opinions), and technological problems (e.g., media recommendation and popularity prediction) which were otherwise hard to address purely by traditional approaches. However, media content on the Internet is unevenly distributed, often depending on platforms, popularity, and bias of Web domains. The connected power where each kind of media can enhance others has not been fully realized. This chapter focuses on two important techniques and three novel applications that utilize the potential of social media by algorithmic detection of trending topics, and utilizing the knowledge learned in developing socially aware multimedia applications by scalable transfer learning. We discuss the recently developed Online Streaming LDA (OSLDA) algorithm for topic mining from social streams and SocialTransfer, a scalable way to transfer information across domains. Using a combination of these techniques, socially aware multimedia applications such as social video recommendation, social popularity prediction, and social query suggestion can be realized.