Sentiment analysis of COVID-19 cases in Greece using Twitter data

Area of Science:

• Computational Social Science
• Public Health Surveillance
• Sentiment Analysis

Background:

• Internet data, including social media and search engine records, has been utilized for epidemic tracking and forecasting for two decades.
• Recent research explores using the World Wide Web to analyze public reactions, emotions, and sentiment during outbreaks, especially pandemics.

Purpose of the Study:

• To assess the effectiveness of Twitter messages (tweets) in real-time sentiment analysis of COVID-19 cases in Greece.
• To correlate public sentiment with the number of COVID-19 cases and data volume.

Main Methods:

• Collected 153,528 tweets from 18,730 users over one year.
• Analyzed tweet content using English-to-Greek translated Vader lexicon and a Greek lexicon.
• Tracked positive/negative sentiment, six specific emotions (surprise, disgust, anger, happiness, fear, sadness), and correlations with COVID-19 cases and tweet volume.

Main Results:

• Surprise (25.32%) and disgust (19.88%) were the dominant sentiments expressed in tweets.
• Sentiment correlation coefficients with COVID-19 cases were weak (Vader: R² = -0.07454; Greek lexicon: R² = 0.167387).
• Sentiment did not significantly correlate with COVID-19 case spread, potentially due to declining public interest.

Conclusions:

• Twitter data can reflect public sentiment, with surprise and disgust being prominent during the COVID-19 pandemic in Greece.
• Public sentiment expressed on Twitter did not show a significant correlation with the real-time spread of COVID-19 cases.
• The study highlights the complexity of linking online sentiment to epidemiological data, suggesting factors like waning public attention influence the correlation.