Análise do nível de subjetividade para detecção de farsas em trechos de notícias falsas.

Abstract

While the easy access, very low cost and rapid dissemination of news through the Internet lead people to consume a huge amount of information every day, it also creates a big issue that is the wide spread of fake news. Although there are tons of works regarding fake news detection, there are very few ones that investigate its structure deeply. Therefore, to better understand how fake news is structured, we propose to consider the subjectivity of news under the premise that the subjectivity levels of legitimate and fake news are significantly different. For computing the subjectivity level of news, we rely on a set of subjectivity lexicons built by Brazilian linguists. We then tagged fake parts of the news articles and split them by falsehood category. After that, we calculated the Word Mover’s Distance (WMD) between these parts and the lexicons to build the subjectivity feature vectors, in order to perform the experiments.We believe that our findings contribute to the progress of studies involving fake news detection.