Online reviews are crucial for identifying the factors that affect the dynamic evolution of tourist sentiment, improving tourist satisfaction. This study employs pre-trained models BERT and BERTopic and social network analysis to examine 228,062 reviews collected from Ctrip using Python. The factors influencing tourist sentiment across natural tourism attractions (NTAs), cultural tourism attractions (CTAs), and theme park tourism attractions (TPTAs) were explored before, during, and after the pandemic. The findings reveal that there was minimal change in the types of factors influencing before and during the pandemic, significant changes in the values of factors during the pandemic, and fluctuations in both the types and values of factors after the pandemic. Regardless of the period, influences on negative sentiment were more dispersed, while positive emotions were more polarized. Based on these insights, we propose theoretical contributions and improvement strategies for enhancing resilience and promoting high-quality development in different types of attractions.