A Novel Two-Stage Framework for User Identification Across Social Networks

Abstract

User identification across social networks refers to the estimation of whether multiple accounts on different social networks belong to the same individual. By integrating user information across multiple social networks, we are able to build a more comprehensive user profile contributing to many applications. Previous work with good performance needs rich user information. However, as the privacy policy changes, it is difficult to obtain such rich information. Besides, the combination ways of multi-dimensional information are also inefficient. Therefore, a novel two-stage user identification framework is proposed. Firstly, we establish user identification model UNM (UserName Based Model) merely using username and TCM (Tweet Content Based Model) only using tweet content information. Then we integrate two models into two-stage user identification framework. Experimental results show that (1) only the use of tweet content information can also achieve good results and (2) two-stage user identification framework can provide promising performance with F-score reaching 97.6%, which proves the superiority of the framework.