Natural language processing analysis applied to COVID-19 open-text opinions using a distilBERT model for sentiment...

Area of Science:

• Natural Language Processing
• Computational Social Science
• Public Health

Background:

• The COVID-19 pandemic significantly impacted global quality of life and lifestyles.
• Government-imposed lockdowns in 2020 necessitated understanding public sentiment during this unprecedented period.

Purpose of the Study:

• To develop and evaluate a natural language processing model for detecting positive and negative sentiments in open-text survey responses from the pandemic era.
• To contribute to the understanding of public feelings during virus-induced lockdowns for future challenges.

Main Methods:

• A distilBERT transformer model was proposed for sentiment analysis.
• Three distinct approaches were employed for model comparison.
• Sentiment analysis was performed on open-text survey answers collected during the pandemic.

Main Results:

• The best-performing model achieved an average Accuracy of 0.823.
• The model demonstrated strong performance with an average Precision of 0.826.
• Average Recall and F1 Score were reported as 0.793 and 0.803, respectively.

Conclusions:

• The proposed distilBERT model effectively detects public sentiment in pandemic-related survey data.
• The findings highlight the utility of NLP in understanding societal responses to global health crises.
• This research provides a foundation for sentiment analysis in future public health emergencies.