LSTM Based Sentiment Analysis Model to Monitor COVID-19 Emotion

Abstract

Psychologists and Social scientists are interested to evaluate how people show their expressions and sentiments about natural disasters, terrorism, and pandemic situations. The covid-19 has raised the number of psychological issues such as depression due to social changes and employment issues. The everyday life of people is disturbed due to the Pandemic situation of covid-19. During the lockdown, people share their opinions on social sites like Twitter and Facebook. Due to this pandemic situation and lockdown, the emotions of people are different, the emotions are categorized as fear, anger, joy, and sad in terms of covid-19 and lockdown. In this paper, we have used machine learning and Natural Language Processing approaches to design an effective machine learning model for the classification of people's emotions related to covid-19. The early detection of sentiment allows for better handling of the pandemic situation and government policies. Text is categorized into fear, joy, anger, and sad sentiment classes. We have proposed a deep learning-based LSTM model for Covid-19 related emotion identification and achieved an accuracy of 71.7% with the proposed model. For the robustness of the proposed model, we considered several machine learning classifiers and compare these classifiers with our proposed model. Data Availability: In this study, an open-source dataset is used:https://www.kaggle.com/code/poulamibakshi/covid-19-sentiment-analysis/data