Nowadays, because of increasing threat of fake news to trustworthiness of online information, recognizing the truthfulness of news can help to minimize its potential problems in society. However, finding truthful information from social media contexts, which covers a large number of subjects, is a very complex task. Fake news detection is more than simple keyword spotting task, the truth of statements cannot be assessed only by context of news, and it is needed to automatically understand human behavior and sentiment in social media that usually are vague and dependent on subject which should be interpreted and represented in different ways.
This talk presents a comprehensive overview of the findings related to fake news. I characterize the state-of-the-art in detection methods. Many of these rely on identifying features of the users, content, and context that indicate fake information. I also study existing datasets that have been used for classifying fake news. Finally, I propose promising research directions for online fake news analysis. At the end I propose an approach to detect fake news with different degrees of fakeness by utilizing machine learning techniques on two levels of learning; unsupervised level and supervised level.