What makes conversations interesting?

摘要: Rich media social networks promote not only creation and consumption of media, but also communication about the posted item. What causes a conversation to be interesting, that prompts user participate in discussion on video? We conjecture people conversations when they find theme see comments by whom are familiar with, or observe an engaging dialogue between two more (absorbing back forth exchange comments). Importantly, is interesting must consequential - i.e. it impact network itself.Our framework has three parts: characterizing themes, participants for determining interestingness measures consequences deemed interesting. First, we detect conversational themes using mixture model approach. Second, determine based random walk model. Third, measure consequence measuring how affects following variables participation related participant cohesiveness diffusion. have conducted extensive experiments dataset from popular video sharing site, YouTube. Our results show our method maximizes mutual information, significantly better (twice as large) than other baseline methods (number comments, number new PageRank assessment).