Summarization of large scale social network activity

This paper presents a novel social media summarization framework. Summarizing media created and shared in large scale online social networks unfolds challenging research problems. The networks exhibit heterogeneous social interactions and temporal dynamics. Our proposed framework relies on the co-presence of multiple important facets: who (users), what (shared concepts and media), how (actions) and when (time). First, we impose a syntactic structure of the social activity (relating users, media and concepts via specific actions) in our temporal multi-graph mining algorithm. Second, important activities along each facet are extracted as activity themes over time. Experiments on real-world Flickr datasets demonstrate that our technique capture nontrivial evolution of media use in social networks.

• Type: Conference
• Year: 2009
• Where: ICASSP
• Authors: Yu-Ru Lin, Hari Sundaram, Aisling Kelliher
• Keywords: ICASSP 2009, Signal Processing, Social, Social Media Summarization, Social Networks