Hybrid Approach to Binary News Classification

Abstract

An abundance of news articles are published every single day. The traditional reader is not equipped with the knowledge to differentiate between articles that are reliable from those that are not. With the onset of news being published by various media outlets through social media, there is considerable concern regarding the reliability of these articles. A well-functioning fake news detection model can help detect potentially unreliable news using a combination of Natural Language Processing and Web analysis. With ample research of the current pre-existing models, our proposed approach applies different mechanics for the same task, while giving enhanced results. The task is divided into two phases. First, to obtain the linguistic features from the article text and second, to analyze the factuality of the data in the article using web analysis, thus ensuring that the semantics and context of the presiding entities referred to in the articles are legitimate. 

Keywords: Fake News, Linguistic features, Natural language processing, Web Analysis, Web scraping.