Detecting Early Signatures of Persuasion in Information Cascades

Abstract : The goal of DESPIC project was developing a technological infrastructure to automatically detect orchestrated campaigns on social media in their early stage of diffusion. Such campaigns include rumors, spread of misinformation, persuasion attempts, and advertising. We designed and implemented a distributed infrastructure for the efficient collection, archival and retrieval of Twitter data, and a framework for clustering messages in topically coherent memes in a streaming scenario. Our infrastructure has served as the basis for the development of machine learning infrastructures to discriminate between naturally trending and promoted content, the identification of social bots, the identification of rumors, and the prediction of burstiness and popularity of memes.