Enhanced Fake News Detection With Natural Language Processing

The ubiquitous use of social media impacts the social life of humans. Fake news negatively affects a person's behavior. Fake news detection in social media has been a challenging problem for the last decade. Researchers are searching for accurate methods to detect fake news as early as possible. This paper discusses two NLP methods, LSTM, and bi-directional LSTM, that detect fake news. This paper uses a fake news challenge dataset, categorized into four stance labels: agree, disagree, discussed, and unrelated. Experimental results showed LSTM model received 92% accuracy in testing time and 95% accuracy in training time. On the extension of the method, Bi- directional LSTM achieved 95% accuracy in testing time and 98% accuracy in training time.