Abstract
This work proposes a deep learning-based efficient approach for fake news detection based on linguistic features in the PolitiFact Corpus, a semi-automatically processed and collected database from the fact-checking website PolitiFact.com, which rates the accuracy of claims by elected officials and others. The input of the machine learning system is first obtained from a text database and then integrated with robust linguistic features obtained based on embeddings of Multidimensional Analysis Tagger (MAT). Then, a fine-tuned deep learning approach (i.e., Attention-based Long Short-Term Memory (LSTM)) is applied to train the features discriminating fake and real news. The trained machine learning model is then utilized later to automatically detect fake news texts. The system has been compared against traditional approaches where the experimental outcomes show the superiority of the proposed approach over them.