Application of the Lightweight BERT Model for the Warning of Tourism Public Opinion Emergencies

The tourism industry, being highly information-intensive and influenced by public opinion, increasingly relies on public opinion monitoring. This study investigates the application of a lightweight Bidirectional Encoder Representations from Transformer (BERT) model in early warning systems for tourism public opinion emergencies. Initially, raw data is preprocessed. Subsequently, the lightweight BERT model is optimized through domain-adaptive fine-tuning and a method combining multi-task learning with model integration. Finally, performance evaluation and robustness testing are conducted. Experimental results demonstrate that the lightweight BERT model, after domain-adaptive fine-tuning and integration with multi-task learning, achieves outstanding performance in early warning for tourism public opinion, with an accuracy of 91%, a recall rate of 89%, and an F1 score of 90%. Compared to other models, the lightweight BERT model exhibits superior performance in terms of accuracy, recall, and F1 score. This study enhances the capability to manage sudden events and public opinion crises in the tourism industry, providing a valuable tool for more effective public opinion monitoring and early warning.