Artificial intelligence for extremist content analysis – FLYER

Nowadays the Internet is a prominent staple in providing youth with learning, entertainment and interactions. Taking advantage of the same dynamics, terrorist groups have begun to reach out to youngsters in an attempt to spread messages of hate and violence or even find potential recruits on the Internet. Bringing together researchers from sociology, linguistic, and artificial intelligence and BERTIN IT, a French software publisher, FLYER aims at creating effective solutions for cyberspace exploration. The project is coordinated by ONERA, an EPIC involved in various security related projects at national and Europe level, and the consortium was built in an effort to combine fundamental research contributions and innovation efforts towards efficient solutions to nowadays problems. FLYER is scheduled to run over two years as a joint research project, and its objective is twofold. First, it addresses the analysis of online extremist content as highlighted by sources of indoctrination, messages and conversations in the French-speaking cyberspace. The goal is to acquire specific knowledge about online extremist content and to shade light on underlying social models and their specific forms of communication. FLYER focuses on radicalized and far right extremism specific content. The second objective is the implementation of innovative solutions for cyberspace exploration. FLYER aims at implementing methods for in-depth characterization of online content, according to domain specific concepts and the subjective engagement expressed by users. In particular, FLYER uses models and resources created by researchers in sociology and linguistics as a solid background to develop procedures for the automatic characterization of digital content (messages, conversations) from lexical, discursive and semantic points of view. For online content filtering and exploration FLYER adopts Adaptive Supervised Learning (ASL), a hybrid paradigm augmenting learning algorithms with adaptation mechanisms acquiring new descriptors, in order to acknowledge and cope with domain evolution. By adopting ASL, the project will provide unprecedented capabilities for a systematic exploration of online content. From a practical standpoint, FLYER aims at offering support to users tackling the propagation of extremist content online: sociology and sociolinguistics researchers can use the tool to detect both relevant website content and ongoing chatting or messaging, to investigate specific patterns of online content, to describe radical conversations and narratives, to build, enrich or validate their models. Intelligence analysts and homeland security experts can use the tool for a more accurate exploration of the cyberspace and for empirical validation of their hypotheses. In addition to breakthrough in research, FLYER can have commercial impact, as the consortium integrates BERTIN IT, a French software publisher which can accelerate the integration of developed functionalities into final products. Thanks to synergies within the consortium, FLYER will deliver several results going beyond the state of art, as well as technological building blocks whose practical usage will reveal new findings on the societal and technological issues at the frontier between real and virtual environments. Results of FLYER can impact the civilian domain, where the project can offer support to mediators of social networks. Military applications are also to be expected, as technological blocks can be used to build intelligence solutions for intelligence and homeland security.