Using Deep Learning Models for COVID-19 Related Sentiment Analysis on Twitter Data
Waqas Haider Bangyal, Saeeda Amina, Rabia Shakir, George Ubakanma, Muddesar Iqbal
Abstract
The significance of sentiment analysis has increased in modern times due to the extensive use of social media platforms as a medium for individuals to express their opinions. Twitter is widely acknowledged as a popular social media site mostly used for microblogging. People often voice their opinions on current events, making it challenging for researchers to accurately classify the mood expressed in these opinions. This research study presents a novel and efficient method for identifying and detecting false or misleading information pertaining to the COVID-19 pandemic. The dataset including artificially created news stories is obtained from a collection of texts and processed via the cycle of natural language processing (NLP). For this research study, three advanced deep learning models were used to predict the emotion of news items, accurately differentiating between genuine and fraudulent ones. This study utilizes Convolutional Neural Networks, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) as deep learning classifiers. Subsequently, it compares the outcomes achieved from these classifiers. The findings suggest that the BiGRU deep learning classifier has exceptional accuracy and efficiency, achieving a remarkable accuracy rate of 91%.