Deep Learning based Fake News Detection For The Kurdish Language

In today’s world, technology has facilitated the publication of fake news on social media. Thus, it becomes very difficult to differentiate real news from fake ones particularly in low-resourced languages such as Kurdish. Therefore, in this study, the author bridges this gap by proposing a hybrid deep learning model named “ACO-CNN” that can effectively recognize false news. The hybrid approach includes the Ant Colony Optimization (ACO) that has been utilized to optimize the Convolutional Neural Network (CNN) hyperparameters to enhance its capabilities to successfully detect fake news. The proposed method was evaluated against the state-of-the-art methods that were incorporated for the same purpose including, Support Vector Machine (SVM), Naïve Bayes (NB), Random Forest, (RF) and Decision Tree (DT) using two existing benchmark fake news dataset namely “KurdFake” and “KDFND”. Our experimental results indicate that the ACO-CNN outperformed other approaches in terms of F1-score on both datasets. This work heavily contributes to the Kurdish Natural Language Processing (NLP) field and the development of effective deep learned based fake news detection tools for under-resourced languages.