Creative Commons

This dissertation was written as a part of the MSc in Data Science at the International Hellenic University. False information can be created and reproduced easily through the Web due to the rapid growth of online social media that enable sharing ideas and thoughts through virtual networks and communities. There are stories that are deliberately fabricated to influence public opinion and cause confusion to the reader. Thus, tackling fake news has become an important issue over the past few years. Several detection approaches have been proposed, such as but not limited to fact-checking and network analysis. However, to the best of our knowledge, most of them can identify fake news after its propagation. Understanding the writing style of fake news could support the early detection of disinformation. In this thesis, we investigate the use of linguistic features for identifying deception in news articles and experiment with state-of-the-art machine learning and natural language processing approaches. Furthermore, we prove that integrating linguistic characteristics into fake news detection models, can improve their prediction accuracy.