Deep Learning Techniques for Fake News Detection

Abstract

Social media has recently become the primary source for people to consume news. Plenty of users prefer to go to social media apps such as Twitter, Facebook, and Snapchat to obtain the latest social events and news. Meanwhile, traditional media is emulating the new media to post their news on the aforementioned apps. This prevalence is a double-edged sword, for the advantage is that users can easily gain access to the news articles they look for on social media. However, it also provides an ideal platform for fake news propagation. The spread of fake news is extremely fast on social media and can cause adverse effects in real life. The unregimented, incomplete censorship and the absence of fact-checking processes make fake news easy to propagate and hard to control. Therefore, fake news detection on social media has become a trending topic that draws tremendous attention, as shown in figure 1. Nevertheless, as pundits dig into the realm of deep learning, some of the studies utilize deep neural networks (DNN) to build frameworks that would help detect fake news. Although impressive progress on the topic has been made, the lack of a review dissertation that summarizes and synthesizes the overall development of the study would be problematic. Hence, this paper aims to summarize different models implemented in recent studies that improve the veracity of fake news detection.