Topic Modeling of Social Media Sleep Health Discourse Using BERTopic: A Study of Naver Blog Big Data

Objectives: This study aimed to identify topics and characteristics of online discourse related to sleep health using social media big data. Methods: We collected 8,002 blog posts from Naver Blog using nine sleep-related keywords (insomnia, sleeping pills, sleep disorder, Stilnox, melatonin, sleep routine, sleep hygiene, overcoming insomnia, can’t sleep) from March to October 2025. Topic modeling was performed using the BERTopic algorithm with kosbert-nli embeddings for Korean text, UMAP dimensionality reduction, and HDBSCAN clustering. Results: Nine major topics were derived from 6,177 documents (77.2%) excluding noise. The topics with the highest proportions were complaints about insomnia symptoms (23.8%) and experiences with sleeping pills (23.7%). Analysis revealed that 85.2% of posts searched with sleep routine were classified under experiences with sleeping pills, indicating high drug dependence. In addition, 63.1% of melatonin posts were discussed similarly to prescription sleeping pills, and 68.3% of sleepless posts were focused on the topic of Stilnox side effects. Conclusions: Online sleep health discourse focuses primarily on pharmacological treatment and highlights the need for health policy interventions to provide information on non-pharmacological approaches and reduce drug dependence. This study demonstrates the potential of social media big data analysis to monitor sleep health information.