A Multifaceted Approach for Identifying Propaganda on Social Networks

Online Social media facilitate human interaction, information sharing, and opinion expression in digital era. Misinformation, disinformation and propaganda propagate quickly through social media platforms and impacts public opinion, political debate, and societal views. Understanding the mechanics of propaganda dissemination and its repercussions on society is critical for reducing its negative implications. Recognizing propaganda will allow individuals to safeguard their values and resist attempts to manipulate them. In this paper we propose a hybrid feature selection technique to classify the propaganda and non propaganda tweets. Data is prepared based on the annotation scheme and features are selected by fused various state of art feature selection techniques. Various machine and ensemble learning classifiers are trained and tested based on the features selected. The results showed that the there is an increase in the performance of all the classifiers trained on the proposed hybrid selection criteria. In future Deep learning techniques may be incorporated to improve the efficiency.