A likelihood-based framework for the analysis of discussion threads

摘要: Online discussion threads are conversational cascades in the form of posted messages that can be generally found social systems comprise many-to-many interaction such as blogs, news aggregators or bulletin board systems. We propose a framework based on generative models growing trees to analyse structure and evolution threads. consider growth determined by an interplay between popularity, novelty trend (or bias) reply thread originator. The relevance these features is estimated using full likelihood approach allows characterise habits communication patterns given platform and/or community. apply proposed four popular websites: Slashdot, Barrapunto (a Spanish version Slashdot), Meneame Digg-clone) article pages English Wikipedia. Our results provide significant insight into understanding how grow have potential applications broader contexts community management design platforms.