Taking the travel commentary information of Xi'an and Chengdu on Ctrip as example, the influencing factors of tourism satisfaction are explored by using word cloud map, social network semantic graph and LDA topic model.Together with the emotion dictionary method to calculate index sentiment value and the analytic hierarchy process to determine index weight, the urban tourism satisfaction evaluation system is constructed based on online text comments. The results show that Chengdu's satisfaction is higher than that of Xi'an. Tourists are more satisfied with the tour guides and itinerary of Chengdu tourism service than Xi'an, while Xi'an has better regional tour features and scenic experience than that of Chengdu.