A Semi-supervised Framework for Misinformation Detection

The spread of misinformation in social media outlets has become a prevalent societal problem and is the cause many kinds unrest. Curtailing its prevalence great importance machine learning shown significant promise. However, there are two main challenges when applying to this problem. First, while much too one respect, misinformation, actually, represents only minor proportion all postings seen on media. Second, labeling massive amount data necessary train useful classifier becomes impractical. Considering these challenges, we propose simple semi-supervised framework order deal with extreme class imbalances that advantage, over other approaches, using actual rather than simulated inflate minority class. We tested our sets Covid-related Twitter obtained improvement F1-measure extremely imbalanced scenarios, as compared classical deep-learning generation methods such SMOTE, ADASYN, or GAN-based generation.