Fake News Detection with Integration of Embedded Text Cues and Image Features

Abstract

A novel approach using Convolution neural Network (CNN) and Long short-term memory (LSTM) has been proposed to find the reliability of the news. In this research, image visual feature with embedded text feature and headline texts have been considered to find the comprehensive results. First the semantic information from the images have been captured as text (news tag line) and this tag has been compared to the original headline text. Individually image and text both are insufficient to find the semantic knowledge of publish news. So, the cosine similarity index (CSI) has been used to predict the reliability of the news. The threshold of CSI has been constrained greater than 0.62 for the news real. A repository has been created named as” imaged fake news”. In this repository 1000 images have been considered with the headline texts, where 367 news were fake and 633 news were real. The accuracy of the proposed method is 91.07%. The result implies that the novel methodology is better than the state-of-the-art method.