Input data for different models.

In March 2020, the outbreak of COVID-19 precipitated one of the most significant stock market downturns in recent history. This paper explores the relationship between public sentiment related to COVID-19 and stock market fluctuations during the different phases of the pandemic. Utilizing natural language processing and sentiment analysis, we examine Twitter data for pandemic-related keywords to assess whether these sentiments can predict changes in stock market trends. Our analysis extends to additional datasets: one annotated by market experts to integrate professional financial sentiment with market dynamics, and another comprising long-term social media sentiment data to observe changes in public sentiment from the pandemic phase to the endemic phase. Our findings indicate a strong correlation between the sentiments expressed on social media and market volatility, particularly sentiments directly associated with stocks. These insights validate the effectiveness of our Sentiment(S)-LSTM model, which helps to understand the evolving dynamics between public sentiment and stock market trends from 2020 through 2023, as the situation shifts from pandemic to endemic and approaches new normalcy.