LSTM Based Approach to Detect Fake News

In the age of technology, every day, we ingest a variety of news data, either willingly or accidentally. People will always use various social media apps and keep looking for information. The information passed through these social media apps is only sometimes valid. Many organizations or individuals spread the news to influence the mass public. This type of information is called fake news. Fake news is a common problem that naturally comes with social media. Detection of counterfeit news with ubiquitous data from social media is almost impossible. Fake news kills real journalism and spreads false information in society. Researchers are still looking for better solutions. In this paper, we find a better solution with Natural Language Processing (NLP). We used fake news challenge datasets for training and testing purposes. We aim to classify the news item into four stance labels: agree, disagree, discuss, and unrelated. We developed a Long Short-Term Memory (LSTM) based solution to organize the news article in the given stance label. We find that our approach classifies data effectively with 99% of accuracy