Fake News Tracker A Tool for Fake News

Fake. News. Tracker: A Tool for Fake News Collection, Detection, and Visualization Kai Shu, Deepak Mahudeswaran, Huan Liu Arizona State University {kai. shu, dmahudes, huan. liu}@asu. edu Introduction • Fake News Detection is an important problem because of its social impacts. • It is challenging because of Ø Lack of adequate labeled data Ø Changing topics of fake news in social media Fake News Detection - Social Article Fusion • Detect using news context and social context Ø Learn News representation using Autoencoder Ø Learn Social temporal engagements using RNN Visualization • Web interface for tweet and feature visualization Ø Data Exploration including trends, geo location, social network and topics Ø Visualizations to compare the feature significance and model performance Proposed Framework Ø Jointly optimize the output of both networks • Contributions: Ø An end-to-end framework for the fake news collection, detection and visualization Ø An ability to collect fake news pieces in streaming manner Ø A detection mechanism using temporal social engagements and news content Experimental Results Proposed model performs better when both news content and social engagements are used. Performance of Fake News Detection 1 0. 9 0. 8 0. 7 0. 6 0. 5 0. 4 0. 3 0. 2 0. 1 0 Accuracy Precision Recall F 1 Accuracy Politi. Fact Dataset Recall Buzz. Feed Dataset SAF /S Data Mining and Machine Learning Lab Precision SAF /A SAF F 1 Future Work • Extend Fake. News. Tracker to establish a Fake. News Repository • Extend Fake. News. Tracker to collect data from different sources or platforms This material is based upon work supported by, or in part by, the ONR grant N 00014 -16 -1 -2257.