We Will Know Them by Their Style: Fake News Detection Based on Masked N-Grams

Thanks to the availability of digital media, users receive daily news reports, opinions and information on a wide variety of topics. These same media allow people to easily share and transmit their own opinions, thus enriching the debate and reflection on topics of public interest. Unfortunately, these circumstances have led to the emergence of fake news to misinform. This phenomenon has reached huge proportions, becoming a serious problem. Different approaches have been proposed to automatically detect fake news, based on analyzing their content, source or dispersion. The objective of the present work is to explore whether the written style of news can be used for this task. The proposed method uses a simple strategy based on keeping the most frequent words while masking the rest. The experiments in four collections, in two languages and different topics, have led to the conclusion that there are common lexical stylistic elements among fake news.