Synonyms-Based Augmentation to Improve Fake News Detection using Bidirectional LSTM

Abstract

Fake news is the news which contains propaganda and not relevant to the actual news. Today, the news in social media are troubling internet user. Hence, a fake news detector is needed to solve the problem. In this research, a fake news detector system based on Recurrent Neural Network (RNN) is developed. The architecture is designed using Bidirectional Long Short-Term Memories (Bi-LSTM) with exploit stance detection for the headline and the body of the news. Evaluation on 50 k news articles from FNC-1 shows that the proposed method produces F1-score of 0.2423 in detecting the fake news.