1. Information propagationanalysis in a socialnetwork site Matteo Magnani* - Danilo Montesi* - Luca Rossi° * University of Bologna, Dept. of Computer Science ° University of Urbino “Carlo Bo”, Dept. of Communication Studies http://larica.uniurb.it/sigsna
10. Data extraction ≃ 10.500.000 posts ≃ 500.000 likes. ≃ 450.000 users. ≃ 15.000.000 di archi (subs). Downloadable from: http://larica.uniurb.it/sigsna/data/
11. Static and dynamicnetwork In otherdisciplinesreconstructing the network is complex. Herewegetnetworkstructures for free. However, the network over which information propagates is verydifferent from the technicalnetwork.
24. Time related to some events Chat: depends on topic more than time. News: the winner takes all. Chat News 7 top commented threads about Mike’s death
26. Research findings 1/2 Users active inside Friendfeed generate much more comments than external users importing their messages into the service. Content production rate follows specific time-trends. The average audience of an entry depends on its posting time with specifically identified trends. Information spreads on High Priority Networks built on top of the technical network. Automated users tend not to generate discussions. The number of comments received by users with more limited entry production rates increases only up to some threshold (information overload).
27. Most conversations have a very quick growth and an evolution that usually ends within a few hours. This is particularly evident for highly commented entries —the presence of many comments often implies a shorter discussion. For informational messages, time is relevant. Given the high rate of answers, an early message may have a saturation effect so that it aggregates the majority of discussions and limits the development of conversations on other similar messages. This does not seem to apply to the second kind of messages, which may start days after the news occurred. Research findings 2/2
28. Moral Identification of some of these factors (source, multimedia, culture, timing, kind of message, active network). Quantitative analysis on a real dataset. The “success” of a post depends on many factors related to its socio-technical context .
29. SIGSNA project (google). Twitter: sigsna matmagnani lrossi Information propagationanalysis in a socialnetwork site Matteo Magnani* - Danilo Montesi* - Luca Rossi° * University of Bologna, Dept. of Computer Science ° University of Urbino “Carlo Bo.” Dept. of Communication Studies http://larica.uniurb.it/sigsna