Recovery of hotels from the crises:evidence from tourists’ emotional changes by deep learning sentiment analysis

The hospitality sector is highly susceptible to crises. Understanding guests’ emotional reactions and attitudes toward hotels during such times is crucial for developing effective retention strategies and revitalizing the industry. This study examines changes in guest sentiment toward hotel attributes during the overlapping crises of the 2019 Hong Kong protests and the COVID-19 pandemic. Using deep learning methods, specifically the BERT language model, the research analyzed 2,941,710 textual units to track sentiment shifts across pre-crisis, crisis, and post-crisis stages. Results indicate significant sentiment fluctuations affecting various hospitality aspects. This research extends deep learning applications in crisis impact assessment and offers strategic insights for hotel managers to craft marketing strategies throughout a crisis lifecycle.