How Many Users Are Enough? Exploring Semi-Supervision and Stylometric Features to Uncover a Russian Troll Farm

摘要: Social media has reportedly been (ab)used by Russian troll farms to promote political agendas. Specifically, state-affiliated actors disguise themselves as native citizens of the United States discord and their motives. Therefore, developing methods automatically detect trolls can ensure fair elections possibly reduce extremism stopping that produce discord. While data exists for some organizations (e.g., Internet Research Agency), it is challenging collect ground-truth accounts new in a timely fashion. In this paper, we study impact number labeled on detection performance. We analyze use self-supervision with less than 100 training data. improve classification performance nearly 4% F1. Furthermore, combination self-supervision, also explore novel features grounded stylometry. Intuitively, assume writing style consistent across because single organization employee may control multiple user accounts. Overall, models based words ~9%