MINING PUBLIC OPINION ON RADICALISM IN SOCIAL MEDIA VIA SENTIMENT ANALYSIS

The identification and classification of hate speech, extremism, and radicalism in social media are very important topics today because they have a wide negative impact on society. Hostile groups use this media to spread their hate speech, ideology, and recruitment of individuals. This study aims to propose a new method using deep learning to classify the utterances of hate, extremism, and radicalism in Indonesian which are posted on social media. This method uses word2vec as word embedding and the combination of Restricted Boltzmann Machine and back-propagation network as our basic classification method. The model can achieve 81.63% accuracy to predict radicalism and hate-speech. Our model outperforms the baseline classifier methods based on the comparison in experimental results.