BACKGROUND: Tobacco control faces significant challenges due to complex narratives that obscure public health messages and the tobacco industry's efforts to spread misinformation to weaken policy initiatives. These challenges complicate sentiment analysis of online content, hindering effective advocacy of tobacco-related discussions.
METHODS: This research introduces an intelligent system to support tobacco control experts in policy advocacy by efficiently classifying and analyzing online media. The system employs advanced techniques to collect Vietnamese datasets and uses a novel fine-tuning of the NS-PhoBERT model, combining the Vietnamese pre-trained PhoBERT framework with Neutrosophic theory to address uncertainty and inconsistencies in information. Applied to 1038 news articles on tobacco company corporate social responsibility (CSR) activities in Vietnam in 2023, the system performs tasks such as classification and sentiment analysis with high accuracy.
RESULTS: Experimental results demonstrate the system’s ability to efficiently screen and analyze large volumes of articles, identifying content that supports the tobacco industry or subtly undermines public health advocacy. This saves significant time and resources, allowing experts to focus on uncovering new insights critical for shaping effective tobacco control policies.
CONCLUSIONS: This study highlights the potential of intelligent systems to transform public health advocacy by equipping experts with advanced tools to combat misinformation and strengthen tobacco control policies efficiently and effectively. By enhancing accuracy and consistency in policy analysis, the system supports real-time monitoring and evaluation, empowering public health organizations to refine strategies, counter misinformation, and advocate for stronger policies. Scalable and replicable, this tool is a valuable resource for future public health campaigns, streamlining analysis processes and ensuring reliable sentiment tracking over time.