Multidimensional analysis of childbirth experiences on Chinese social media: an integrated approach using LDA, sentiment analysis and deep learning.

Objective: Childbirth is a major life event with complex physical and psychosocial consequences. This study examined how women's childbirth experiences are constructed in Chinese social media narratives and identified key themes and emotional patterns across different modes of delivery. Methods: We conducted a cross-sectional content analysis of 17,783 childbirth-related posts from Weibo and REDnote. Latent Dirichlet allocation (LDA) was used to extract topics, and a pre-trained Chinese BERT model was fine-tuned for three-class, sentence-level sentiment classification; model performance was evaluated using accuracy, F1-score and AUC-ROC. Results: LDA identified six main themes, with "vaginal birth process and pain experience" and "emotional meanings and family value of childbirth" accounting for the largest proportions. Negative emotions substantially outweighed positive ones and were particularly concentrated in vaginal birth narratives, whereas caesarean section and postpartum recovery posts were more neutral and informational. The integrated LDA-BERT approach achieved good performance (AUC-ROC = 0.89), supporting the robustness of sentiment estimates. Conclusion: Childbirth narratives on Chinese social media present a predominantly negative yet "bittersweet" emotional profile, with marked differences between vaginal and caesarean experiences. These findings underscore the need for tailored perinatal interventions, including individualized labour pain management, strengthened psychological support and more nuanced online health communication.