Detecting Conspiracies on Twitter About COVID-19 Using Deep Learning Techniques

Detecting Conspiracies on Twitter About COVID-19 Using Deep Learning Techniques by Faiza Rasheed (2022)

With the increase in use of social media, the dissemination of propaganda and conspiracies is also growing sharply around the world in many aspects of life, such as the economy, politics, environment, health, among many others. Many users of social media blindly follow and become a part of the propaganda without verifying its authenticity. Recently, a number of conspiracies related to Coronavirus Disease 2019 (COVID-19) got popular and had a negative impact on the health of millions of people. False information is considered one of the most severe threats to the world’s economy and civilization, and it has the potential to cause enormous economic and social damage to societies. People use Twitter to discuss a wide range of global topics, as well as share their own thoughts and opinions with others across the world. Over the past few years, the spread of conspiracies via social media regarding COVID-19, specifically raised a serious concern owing to the disastrous effect it has on communities around the world in the terms of economy, health, politics, and also in handling such a devastating pandemic. COVID-19 is one of the worst pandemics in the history of humankind. However, uncertainty arose over handling of COVID-19 due to the proliferation of several conspiracy theories around it. It affected public opinion and made it more difficult to properly gather information on how to deal with the pandemic systematically. Stopping conspiracy theories is a challenging task. Identifying conspiracy theories would be the first step in this direction. It can be achieved by employing state-of-the-art language processing techniques. In this research, we use Machine Learning and Deep Learning techniques to fill-in the current knowledge gap in the endeavor to build a more effective method for identifying conspiracies on Twitter. We have conducted experiments on tweets related to conspiracy theories about COVID-19. In this research, we find out which kind of machine learning (ML) or deep learning (DL) algorithm is most suited for identifying conspiracies in tweets about COVID-19. The KNN, RF, MNB, LR and L-SVC are some of the machine learning approaches used in studies. Deep learning algorithms such as CNN and RNN-LSTM are also used. According to the findings of this study, deep learning with the BERT-CNN model outperforms other deep learning and machine learning algorithms, regardless of whether the data was balanced or imbalanced. The random forest model continues to perform as the second-best option among the nine models available for examining potential conspiracies about COVID-19.