Fake news classification is designed to classify the news from the datasets as true or fake news. The necessity of a fake news classification system is due to the rapid spread of fake news through the internet. Traditional machine learning models are not suitable for fake news classification due to hand-crafted feature extraction from the input textual content.
Deep learning models are the widely used data-driven automatic fake news detection method. However, Deep learning approaches require additional word embedding representations to analyze the correlation between the news and its surrounding factors in the content. Contextualized word embedding representation provides efficient word embedding representation of input data that helps understand the relationship between the news and its context. Fake news classification based on deep contextualized word representation effectively categorizes the news as real or fake.