Abstract
As the economy and society develop and the standard of living improves, people's health awareness increases and the demand for health information grows. This study introduces an advanced BERT-LDA model to conduct topic-sentiment analysis within online health communities. It examines nine primary categories of user information requirements: causes, symptoms and manifestations, examination and diagnosis, treatment, self-management and regulation, impact, prevention, social life, and knowledge acquisition. By analyzing the distribution of positive and negative sentiments across each topic, the correlation between various health information demands and emotional expressions is investigated. The model established in this paper integrates BERT's semantic comprehension with LDA's topic modeling capabilities, enhancing the accuracy of topic identification and sentiment analysis while providing a more comprehensive evaluation of user information demands. This research furthers our understanding of users' emotional reactions and presents valuable insights for delivering personalized health information in online communities.