Stance Based Fake News Detection for Amharic news Using Dense Neural Network

Abstract

Fake news is characterized as a story made up with the deliberate of misdirecting or deluding. In this article we display the arrangement to the fake news location action utilizing Profound Learning structures for the Amharic dialect. Gartner's investigate [21] predicts that "By 2022, most individuals in developed economies will devour more untrue data than genuine data." The exponential increment within the generation and dispersion of wrong news in Ethiopia and within the world presents the quick have to be consequently tag and identify such bent news articles. In any case, programmed discovery of fake news may be a troublesome assignment to achieve because it requires the demonstrate to get it the subtleties of common dialect. In expansion, most existing fake news discovery models treat the issue in address as a twofold classification movement, which limits the model's capacity to get it how related or irrelevant detailed news is compared to genuine news. To address these gaps, we present neural network architecture to accurately predict the position between a given pair of titles and the body of the article in Amharic language. Our model able to achieve 95.21% accuracy on test data