The Words of Radicalization: Developing a Bag of Words for Their Implementation Into a Classifying Machine Learning Model

• In Event: Towards a Comprehensive Understanding of Radicalization: Assessment of the Phenomenon, Prevention Tools and Counter-radicalization Policy Analysis.

Abstract

It is known that there are events that generate massive conversations on Twitter, as well as numerous messages are produced with violent and radical content. However, due to the nature of Twitter’s environment, the big amount of data makes them virtually unmanageable for law enforcement agencies.

In this sense, the main aim of the study is the development of a bags of words which contains: a) radical terms and b) non-radical terms but closely related to radical discourse. To achieve this aim, a set of data bigger than 250,000 tweets published in spanish after the attacks on Charlie Hebdo and another sample bigger than 270,000 tweets referring to the 13-N Paris bombing have been used. Based on the taxonomy of violent communication and the online hate speech of the main author, radical and violent tweets were pragmatically detected and classified. Later, cluster analysis of the tweets was made establishing the semantic relations between radical and non-radical words associated.

Authors

• Fernando Miró Llinares, CRIMINA Research Center for the Study and Prevention of Crime (Miguel Hernández University)

• Francisco Javier Castro Toledo, Miguel Hernández University