Fake news detection in Slovak language using deep learning techniques

In recent years, the spreading of fake news presents a serious issue in the online environment. Automatic methods able to identify them from the text are being massively explored and deployed on social platforms and online media. Such detection methods are based on a combination of natural language processing and machine learning techniques. Deep learning became a very popular choice in many text processing tasks, fake news detection included. Numerous studies apply the advanced deep learning models to detect fake news and related phenomena from the English text. This paper focuses on the detection of fake news from the news articles written in the Slovak language. To successfully train deep learning models, we created a labelled dataset consisting of the political news articles published by online news portals as well as suspicious conspiratory portals. We trained two architectures, CNN and LSTM neural networks using this data. The performance of the models was experimentally evaluated using standard classification metrics.