Local COVID-19 Severity and Social Media Responses: Evidence From China

Area of Science:

• Social Sciences
• Computational Social Science
• Public Health

Background:

• The COVID-19 pandemic caused widespread societal disruption and economic challenges.
• Social media platforms became crucial for public expression and information exchange during lockdowns and behavioral regulations.
• Understanding public sentiment is vital for navigating pandemic-related socioeconomic issues.

Purpose of the Study:

• To analyze the evolution of public moods expressed on social media concerning COVID-19.
• To empirically test the correlation between pandemic severity and public sentiment.
• To develop an explanatory pipeline for sentiment classification and socioeconomic issue resolution.

Main Methods:

• Utilized Sentiment Knowledge Enhanced Pre-training (SKEP), a state-of-the-art NLP model, to label Sina Weibo tweets from 2020.
• Employed random forest and linear probit models for feature explanation of the sentiment prediction.
• Incorporated geo-economic control variables to analyze the relationship between local COVID-19 severity and sentiment.

Main Results:

• Demonstrated a significant evolution of moods on social media related to COVID-19.
• Empirically confirmed a strong negative linear relationship between local COVID-19 severity and local sentiment.
• Identified key words influencing sentiment prediction through feature importance analysis.

Conclusions:

• Pandemic severity directly influences public sentiment expressed on social media.
• The study provides a practical and interpretable framework for sentiment analysis in socioeconomic contexts.
• Findings offer valuable insights for policymakers addressing public well-being during health crises.