Leveraging Natural Language Processing to Analyse the Temporal Behavior of Extremists on Social Media

Abstract

Aiming at achieving sustainability and quality of life for citizens, future smart cities adopt a data-centric approach to decision making in which assets, people, and events are constantly monitored to inform decisions. Public opinion monitoring is of particular importance to governments and intelligence agencies, who seek to monitor extreme views and attempts of radicalizing individuals in society. While social media platforms provide increased visibility and a platform to express public views freely, such platforms can also be used to manipulate public opinion, spread hate speech, and radicalize others. Natural language processing and data mining techniques have gained popularity for the analysis of social media content and the detection of extremists and radical views expressed online. However, existing approaches simplify the concept of radicalization to a binary problem in which individuals are classified as extremists or non-extremists. Such binary approaches do not capture the radicalization process's complexity that is influenced by many aspects such as social interactions, the impact of opinion leaders, and peer pressure. Moreover, the longitudinal analysis of users' interactions and profile evolution over time is lacking in the literature. Aiming at addressing those limitations, this work proposes a sophisticated framework for the analysis of the temporal behavior of extremists on social media platforms. Far-right extremism during the Trump presidency was used as a case study, and a large dataset of over 259,000 tweets was collected to train and test our models. The results obtained are very promising and encourage the use of advanced social media analytics in the support of effective and timely decision-making.