Detecting Subgroups in Online Discussions by Modeling Positive and Negative Relations among Participants

摘要: A mixture of positive (friendly) and negative (antagonistic) relations exist among users in most social media applications. However, many such applications do not allow to explicitly express the polarity their interactions. As a result research has either ignored links or was limited few domains where are expressed (e.g. Epinions trust/distrust). We study text exchanged between online communities. find that can be predicted with high accuracy given they exchange. This allows us build signed network representation discussions; every edge sign: denote friendly relation, an antagonistic relation. also connect our analysis psychology theories balance. show automatically networks consistent those theories. Inspired by that, we present technique for identifying subgroups discussions partitioning singed representing them.