AI techniques to analyse a social network on text, user and group level: application on Galaxy2

Online Social Networks have taken a huge place in the informational space, and are often used for advertising, e-reputation, propaganda, or even manipulation, either by individuals, companies or states. As the quantity of information makes the human exploitation difficult, solutions to support the decision makers can only come from the use of AI techniques to extract intelligence from posted messages, to qualify the user behaviours, and to identify the social structure. In this article, we illustrate how to exploit such techniques on a very peculiar social network, named Galaxy2, hidden in the Dark Web. We propose an analysis of 1000 days of activity using NLP techniques to find the most interesting topics and to discover key actors. We then proceed with a MLbased profiling of the user behaviours. Finally, we introduce influence and cohesion scores for groups of users, which help their characterisation and evaluation.