Real and Fake News Classification Using Data Science Process

Abstract

Because of the harm it does, the identification of fake news has piqued the interest of a slew
of academics. Semantics of news articles is the primary focus of most available methods for
spotting false news. However, when the substance of the news is brief, the detecting
performance will substantially decline. Fake news detection multi-task learning (FDML) is
introduced in this research based on the following findings: For example, certain themes have
a larger proportion of bogus stories, and some journalists are more likely to produce phoney
stories than others. For brief fake news, the FDML model studies the influence of topic labels
on the detection performance and introduces context data of news at once to improve
detection performance. Representation and multi-task learning components comprise the
FDML model in order to train the false news detection task, as well as the classification
problem, concurrently. Until now, this has been the only attempt to combine the two methods
of detecting bogus news. On a real-world fake news dataset, the FDML model outperformed
state-of-the-art approaches, according to the findings of the experiments.