Slide 1

A measurement-driven Analysis of Information Propagation in the Flickr Social Network Meeyoung Cha Alan Mislove Krisnna P. Gummadi Slide 2 Whats Flickr? Flickr is an image and video hosting website, web services suite, and online community platform.imagevideo hostingwebsite web servicesonline community As of November 2008, it claims to host more than 3 billion images. Slide 3 Why do research on it? Online social networking websites have been exploited as a platform for the viral marketing of content, products, and political campaigns. major movie studios place trailers for their movies on MySpace; US presidential candidates ran online political campaigns on YouTube; individuals and amateur artists promote their songs, artwork, and blogs through these sites. Other reasons are in conclusions. Slide 4 The assumption is --- It is widely believed that user-to-user exchanges, also known as word-of-mouth exchanges, can spread content, ideas, or information widely and quickly throughout the network. Slide 5 Why do research on it? April 25April 29 Slide 6 Why do research on it? But, hang on.. This assumption is incorrect here Slide 7 Why do research on it? neither the characteristics of information propagation in social networks nor the mechanisms by which information is exchanged are well understood. So lets find out the truth. Slide 8 Whats all about in this paper? (a) how widely does information propagate in the social network? Not widely within two yards (b) how quickly does information propagate? Not quickly, it takes a long time (c) what moves information? Social cascades Slide 9 Biased dataset? NO 1. A partial crawling only affects in-degree and out- degree. 2. To capture the dynamics of friends relationships, the authors conducted the crawling once per day. A comparison: Meeyoungs : 2.5 million users, 33 million links, 34 million favorite markings over 11 million distinct photos Mine: about 38,000 pages of NU website 38,000 * 10 = 380,000 *10 = 3,800,000 Slide 10 Picture popularity different metrics possible for ranking pictures; they can be ranked based on their popularity in terms of the number of views the number of comments the number of fans Slide 11 Why the number of fans ? 1. correlation coefficient The number of views is not strongly correlated with the number of comments (0.13) or the number of fans (0.23). On the other hand, the number of comments pictures receive is highly correlated with the number of fans (0.60). Kill views 2. comments vs. fans Comments could be either way, good, bad. Fans: book mark it, because users love it. Kill Comments Slide 12 Propagate widely? 2 tricky approaches 1.overlap between global hostlist of pictures and local hostlist of pictures. 2.Distance from fans to picture owners. Slide 13 Propagate quickly? Both for popular and less-popular photos. First, in aggregate, many photos do show an active rise in popularity during the first few days after they are uploaded. Second, after the first few (10-20) days, most pictures, in aggregate, enter a period of steady linear growth. For a majority of pictures, over 40% of fans were acquired after the first 100 days Slide 14 What moves information? Featuring Search results Links between content External links Social network Biased? No Slide 15 What moves information? 1.53% of all favorite markings were propagated through social links 2.studying other metrics influence requires a richer data set and is beyond the scope of this paper. External links. 3.Many features, using one, shed light on, enough. CDN case Slide 16 Social cascades move information fast? The average delay in information propagation across social links is about 140 days. Conduct this experiment on pictures with more than 100 fans. Biased? No Slide 17 Less popular photos do not have lots of fans. Do not spread over the whole network There is no need to investigate the question : how long does it take for these less popular photos to propagate along a social link? Slide 18