Enhancing Public Health Feedback Analysis

Abstract

This thesis focuses on implementing text mining techniques, to analyse public comments on Dutch healthcare policies. With the emergence of COVID-19, there was a significant increase in public anxiety and uncertainty, leading to a surge in data from various communication channels. This research consists of two parts, identifying questions, doubts, and concerns within these comments using text classification and identifying topics using topic modelling approaches. The study evaluates the effectiveness of different topic modelling techniques like Latent Dirichlet Allocation (LDA) and BERTopic. To add to it, classification methods, including a rule-based approach, Naive Bayes, logistic regression, and DistilBERT are also implemented. The findings showed that advanced models like BERTopic and DistilBERT provide more nuanced and accurate insights into public sentiment, thereby aiding policymakers in responding effectively to public feedback. This research has broader implications for enhancing public health communication strategies and can benefit other governmental institutions globally.