Fake News Detection using CNN and Attention

Project description: The explosive growth in fake news and its influence on democracy, justice, and public trust has increased the demand for fake news analysis, detection and intervention. In this project, I used text of news, alongside with news’ metadata such as speaker, speaker’s job, and party for fake news detection. I proposed a hybrid convolutional-attentive model that takes use of text statements and meta-datas. The proposed model doesn’t treat all parts of news uniformly but it tries to focus on the important parts of the text statements to detect fake news. I evaluated the model by reporting accuracy, f1, and other metrics and showed that the proposed model can slightly enhance the performance of the baseline model. For more details see report and source code.