Fake News Detection in Social Media Using Deep Learning Techniques

Abstract

Social media platforms have made it simple and quick to share information, newlinewhich has increased the transmission of news and content among millions of users. newlineHowever, this ease of use has also made it simpler for false information to spread newlinerapidly that has a very significant impact. The low-quality false news affects the newlinepeople s confidence about the society. The negative impacts of fake news extend far newlinebeyond just individuals and society. They can also cause significant harm to newlinebusinesses and governments. False news involves spreading of false information for newlinepolitical or financial gain. Online social networks detect fake news challengingly newlinebecause fake news data is lacking and fake news is written by human. False news newlinedetection systems face many difficulties, including extensive training timeframes, newlineoverfitting during training, overthinking, and difficulty in discerning real news from newlinefake and large dimensional data. In addition, the detection of false news is one of the newlineNatural Language Processing (NLP) tasks on Deep Neural Networks (DNNs) that newlinerequires a significant amount of processing power during training. newlineManual fact-checking is unable to keep up with the enormous amount of data newlineproduced by social media networks. This leads to the development and application of newlineautomated techniques for detecting false news. This research addresses the existing newlinechallenges to develop a robust, reliable and effective approach which leverages each newlineunique model epoch to achieve higher and more stable performance for fake news newlinedetection system. newlineThe proposed attention-based false news detection system using deep neural newlinenetwork models like ACNN, ALSTM, and ABiLSTM focuses on most important newlineparts of the text. These models ability to highlight relevant words or phrases, which newlinehelps in better understanding context and nuances that are often vital in distinguishing newlinefake news from the real news. These models are allowed to filter out unnecessary or newlinemisleading information, making them more resilient to noisy data.