Leveraging AI-powered Sentiment Analysis for Enhancing Risk Communication in Radiological Emergencies

Radiological emergencies require clear, credible, and empathetic communication to support public safety and decision making. This paper explores the use of sentiment analysis tools - commonly used in commercial and social media contexts - to evaluate and refine radiological risk communication messages. Twelve pre-trained sentiment analysis models, including rule-based and transformer-based approaches, were applied to sample messages from Federal radiological communication guides. Results demonstrate that transformer models, particularly those with emotion classification capabilities, offer nuanced assessments of message tone. The findings suggest that sentiment analysis tools can support message development by improving clarity, emotional resonance, and alignment with risk communication principles. Recommendations are provided for future work and model fine-tuning specific to radiation-related terminology and the domain of risk communication for radiological emergencies.