Arabic Fake News Detection Using Deep Neural Network Transformers

Over the past few years, fake news in Arabic language has spread widely. This spread has presented serious problems and challenges for people, societies, and governments such as eroding public confidence in the media and shaping public opinion. Many fake news detection models have been introduced. However, the majority of them do not take into account the recent advancement in Natural Language Processing (NLP), particularly the transformers. In order to fill this gap, we introduce a Transformer-based Arabic Fake News Detection Using Deep Neural Network architecture that analyzes and classifies Arabic news articles based on their credibility. We used an extensive dataset of Arabic fake news called (AFND) to train the proposed model. Then, we validate the proposed model against unseen data samples. The experimental results show a high superiority in classification accuracy (approximately 73.7%), which demonstrate the effectiveness and efficiency of the proposed model in distinguishing between authentic and fake news. The results highlight how crucial to utilize cutting-edge machine learning techniques such as transformers in fake news detection systems to enhancing their reliability, especially for Arabic language.