Detecting Toxic Content using Open Source Social Media: An Actor-Centric Approach

This experiment used the Twitter profile of a U.K. national, verified to be an active foreign fighter (FF) in the Syrian civil war, as the seed for constructing a network topology based on social media interactions. After successful seed account verification, data collection was initiated using Twitter’s public API. Radical influence was assumed to extend up to three degrees of separation from the seed profile. However, experimentation with data collection parameters showed that data from the top 30 most communicated-with accounts, and up to two degrees of separation, was optimal.  The social media interactions captured by the two-hop snowball sample were used to model a social network topology. Modularity of the nodes in the resultant network was then used to identify other Twitter users as members belonging to the online Syrian FF community.