Content-based video retrieval has been a very efficient technique with new video content, but it has not regarded the increasingly dynamic interactions between users and content. We present a comprehensive survey on user-based techniques and instrumentation for social video retrieval researchers. Community-based approaches suggest there is much to learn about an unstructured video just by analyzing the dynamics of how it is being used. In particular, we explore three pillars of online user activity with video content: 1) Seeking patterns within a video is linked to interesting video segments, 2) Sharing patterns between users indicate that there is a correlation between social activity and popularity of a video, and 3) Editing of live events is automated through the synchronization of audio across multiple viewpoints of the same event. Moreover, we present three complementary research methods in social video retrieval: Experimental replication of user activity data and signal analysis, data mining and prediction on natural user activity data, and hybrid techniques that combine robust content-based approaches with crowd sourcing of user gener- ated content. Finally, we suggest further research directions in the combination of richer user- and content-modeling, because it provides an attractive solution to the personalization, navigation, and social consumption of videos.


Chorianopoulos, K., Shamma, D.A., and Kennedy, L. 2013. Social Video Retrieval: Research Methods in Controlling, Sharing, and Editing of Web Video. In: N. Ramzan, R. van Zwol, J.-S. Lee, K. Clüver and X.-S. Hua, eds., Social Media Retrieval. Springer, 3–22.   BibTeX