Mapping the "X" Debate: Water Fluoridation Sentiment Analysis With Advanced Machine Learning.

Objectives: This study aimed to examine public sentiment regarding community water fluoridation (CWF) using data from "X" (formerly Twitter) over the past decade. The goal was to understand public opinion trends and identify opportunities for targeted public health communication.

Methods: We conducted a sentiment analysis utilizing a natural language processing technique. Specifically, we applied the Sentiment Intensity Analyzer tool to classify tweets related to CWF into negative, positive, or neutral categories. Additionally, a word co-occurrence network analysis was performed to explore key discussion themes. We also compared machine learning models to assess their accuracy in classifying tweet sentiments.

Results: Analysis of the tweets revealed a balanced distribution of sentiments: 37.4% negative, 34.4% positive, and 28.2% neutral. Peaks in public engagement occurred between 2015 and 2016, with a subsequent decline after 2018. Sentiment spikes were often associated with significant events, including policy changes and media coverage. The word co-occurrence network highlighted recurring themes related to safety and dental health. Among the machine learning models evaluated, Logistic Regression demonstrated the highest accuracy in sentiment classification.

Conclusions: Our findings highlight the polarized nature of public sentiment toward CWF and the temporal fluctuations in public engagement. These insights can inform public health policymakers in developing timely, targeted communication strategies. Specifically, efforts to engage neutral audiences through transparent messaging and counter misinformation during key periods may strengthen public trust in CWF. The application of machine learning in this analysis underscores its value in enhancing real-time monitoring and supporting evidence-based public health strategies.