Public participation serves as a critical component of rural health interventions and epitomizes the full realization of people’s democracy. Consequently, it is essential to tailor rural health construction based on public feedback. This study introduces an opinion mining model based on Long Short-Term Memory (LSTM) networks, designed to extract public opinions from intelligent media platforms. The methodology includes data preprocessing through text filtering, word segmentation, and lexical tagging to prepare the data for analysis. To enhance the model’s performance and avoid overfitting, dropout techniques were employed during training. Opinion classification was subsequently performed using a softmax function. Initial findings from the opinion mining process indicated that 38.29% of the analyzed comments expressed a negative view of rural health conditions. Following targeted interventions to address areas receiving low sentiment scores, a notable improvement in perceptions was observed. Specifically, the sentiment score concerning the attitudes of healthcare workers in the village increased by 14.75%. Additionally, enhancements in waste management practices led to a 19.34% increase in the related sentiment score, contributing to an overall rise of 19.85% in positive public sentiment. These results underscore the efficacy of employing this LSTM-based opinion-mining approach in fostering improvements in rural health environments through informed public participation.