Taking 27 urban parks in Kunming as an example, post use evaluation texts from social network platforms is collected . Based on content analysis method, ROST Content Mining and Excel software are used to conduct high-frequency word statistics, semantic network analysis, emotional tendency and statistical analysis on the text data. Qualitative and quantitative methods are combined to summarize and classify the perception factors of recreational activities in urban parks in Kunming, Mining hidden interrelationships and distribution patterns in review data. The results show that people's attention to different functional types of parks and their constituent elements varies. Finally, based on the true feelings of tourists, suggestions for improvement are proposed from three aspects, including environmental planning and design, cultural landscape creation, and service facility level.