Merging deep learning model for fake news detection

Fake news attracted attention both from the public and the academic communities and represents a phenomenon that has a significant impact on our social life, especially on the political world. Further more, fake news phenomenon provide an opportunity for malicious parties to manipulate public opinion and events such as elections. In this work, we propose a merged deep learning model that detect fake articles regarding different characteristics. Therefore, we use word embedding technique and convolutional neural network to extract text based features and compare different architecture of deep learning while merging two CNNs with different metadata (Text, title, and author). We show on real dataset that the proposed approach is very efficient and allows to achieve high performances.