Integration of NLP and Deep Learning for Automated Fake News Detection

Abstract

In today’s digital era, the widespread issue of fake news significantly undermines the credibility of information, highlighting the need for sophisticated detection methods. This study explores into the use of Natural Language Processing (NLP) and Deep Learning to automate the identification of fake news. The study evaluates several neural network architectures, including Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), Bidirectional LSTM, Artificial Neural Networks (ANN), and a combined CNN+Bidirectional LSTM model. Our extensive testing revealed that the hybrid CNN+Bidirectional LSTM model stands out, achieving an impressive accuracy of 98.13%. This model effectively merges the convolutional layers that extract features with bidirectional LSTM layers that capture temporal dependencies, thus improving its ability to detect subtle textual indications of fake news. The study advances the field of fake news detection by empirically demonstrating the superior performance of specific model combinations, offering crucial insights for developing more robust and precise systems to tackle the ongoing issue of misinformation on digital platforms.