A bimodal longitudinal investigation on changes in sentiments over social media interactions owing to COVID-19 pandemic.

The shift in human sentiments across cyberspace during a pandemic was unprecedented before COVID-19, as no prior pandemic had occurred alongside an interactive online environment. In response, we conducted a comprehensive bimodal longitudinal analysis of sentiment trends throughout the COVID-19 pandemic, utilizing a dataset from 569 Twitter users over 724 days spanning from 2019 to 2020. Our study involved examining 56,789 Tweets, focusing on sentiment shifts in both textual content and images. We further investigated existing sentiment classifier libraries and their potential integration to develop a novel classification technique aimed at enhancing sentiment analysis in text-based Tweets. Additionally, we carried out exploratory data analysis on the collected sentiments and images to identify patterns of sentiment changes attributed to the COVID-19 outbreak as expressed on social media. Our findings indicate a 10.55% increase in negative text sentiment and a 24.52% decrease in positive image sentiment during the pandemic. The bimodal investigation reveals a correlation between text and image sentiment expression, highlighting changes over the two years covering both pre-pandemic and pandemic phases, and identifying change-points for each sentiment type during different phases. This study introduces a novel framework for bimodal sentiment analysis and offers valuable insights for policymakers and social scientists in evaluating public sentiment during global crises.