social media became a fertile soil for various threats, extremism, and radicalization. This challenged policy-makers, researchers and practitioners. Preventing such extreme activities from happening becomes an ultimate priority at local and global scale. This paper introduces a new intertwine between radicalization and natural language processing capable of estimating the risk score of individuals based on their social media activities. The system uses a hybridized ERG22+ and VERA-ER model, which classifies individuals as high or low risk radicalization profile. The developed system was tested and validated on the Video Comments Threat Corpus dataset and Twitter pro-ISIS fanboys datasets where it achieves 95.1% and 64.9% accuracy, respectively.