Fake News Detection Using LDA Topic Modelling and K-Nearest Neighbor Classifier

The spread of the COVID 19 virus has dramatically impacted global society by modifying its lifestyle. Social networks, video streaming tools, virtual collaborative environments have been the primary source of communication through the Internet. This suspension of the “real” has led all activities to be declined through new places and contexts of virtual discussion, increasing new problems, including the most important related to the spread of so-called Fake News. The spread of such news can be devastating: consider what is happening during the critical vaccination phase for COVID 19. In this scenario, systems able to recognize, in a practical way, the truthfulness of news are becoming more and more valuable. This paper aims to present an approach that combines probabilistic and machine learning techniques such as Latent Dirichlet Allocation and K-NN in combination with Context-Awareness techniques to identify the veracity of the news. Adopting Context-Awareness techniques within the proposed system allows a better definition of the operational context Fake News refers to, reducing the problems of semantic polysemy. The first results obtained through standard datasets or using data from real contexts are very interesting and promising.