Bilingual Fake News Detection: Leveraging Deep Learning for Arabic and English Media

Nowadays, the news through this mass technology has a great impact on the information we share, and with this huge number of users, platforms have become a generator of a huge amount of news. They may contain claims such as political propaganda or aggressive content like bullying or harassment phrases. Sharing such fake news results to many struggles like hate speeches and aggressive actions in society so identifying fake and real news is a necessary requirement. Providing such a requirement in multiple languages is important and in this research, our aim is to develop a system able to detect fake news in both Arabic and English language based on textual features. The difference between the two datasets is the preprocessing phase Arabic needs more preprocessing due to the punctuation. This paper uses deep learning and machine learning algorithms for the detection of fake news in Arabic and English. For the English dataset, machine learning models such as gradient boost, logistic regression, decision tree, and random forest were employed, along with deep learning models like GAN, RNN, and LSTM. We employed random forest, decision tree, and logistic regression as machine learning models and LSTM, GAN, and RNN as deep learning models for the Arabic dataset. According to experimental results, the Arabic dataset yields the highest accuracy of 85% for LSTM and GAN, while the English dataset yields the highest accuracy of 99% for gradient boosting