This paper proposes an identification framework for dynamic risk perception with “Questions & Answers (Q&As) + travel notes”, which newly attends to the dynamic nature of risk perception and overcomes the liabilities of traditional data collection methods, such as questionnaires and interviews, which induce high costs in data acquisition, tend to produce small sample sizes and suffer from large sample deviations. Via 2627 Q&As released by tourists before travel and 17,523 travel notes released by tourists after travel, the dynamic change in 20 identified risks before and after travel to Tibet is portrayed with the help of text mining technologies, which can automatically identify risk perception types and sentiment tendencies from massive amounts of textual data. The study finds that before travel, tourists usually underestimate risks related to safety, health and time but overestimate risks related to transportation, route selection and season. The results of the study are not only informative for destination tourism risk management and image promotion but also important for tourists to form more reasonable risk assessments.