VARIATIONAL AUTOENCODER FRAMEWORK FOR MULTIMODAL FAKE NEWS DETECTION

Abstract

Online Social media for news utilization is a have its pros and cons. If we ponder on the positives outcomes for this, it includes easy access, negligible cost, smart categorization and outreach to the very customer in seconds. But, as every coin has two sides and when we flip side of this, a series of issues come up which need immediate attention and most important among them is spreading of fake news. This has become a serious threat for the governments of countries to keep their harmony intact, keep faith of public in democracy and justice and sustenance of public trust. Subsequently detection in fake news, especially in web based platform has become a rising examination topic of interest that is pulling in colossal consideration. Current set of detection algorithms are specially indicating their powerlessness to gain proficiency with the mutual portrayal of text and visuals joined (popularly known as multimodal) information. Therefore, we present a variational auto encoder based framework, which consists of three major components encoder, decoder and fake news detector. It utilize the concatenation of visual latent features from three popular CNN architecture(VGG19,ResNet50,InceptionV3) combined with textual information to detect fake news with the help of binary classifier. We have shown the investigation on two publically available Twitter dataset and Kaggle dataset. The experimental result shows that out model improves state of the art method by the margin of ~2% in accuracy and ~3% in F1 score.