A Closer Look at Fake News Detection: A Deep Learning Perspective

Abstract

The increasingly rapid pace of spreading fake news is considered a problem in conjunction with the increasing number of people who are relying upon social media to get news. That earns widespread attention from research communities due to the negative impact and influence of fake news on public decisions. Consequently, the current research strives to illuminate on fake news problem and the process of detecting fake news using deep learning approaches. Using the Fake News Challenge (FNC-1) dataset, we have developed different models to detect fake news based on the relation between article headline and article body. Our models are assembled mainly from Convolutional Neural Network (CNN), Long Short-Term Memory network (LSTM) and Bidirectional LSTM (Bi-LSTM). In the contrary of other studies on the same dataset where they reported accuracy for a test data derived from the same training dataset, our experiments achieved 71.2% accuracy for the official testing dataset.