Construction and implementation of knowledge enhancement pre-trained language model for text sentiment analysis

With the in-depth development of natural language processing technology, text sentiment analysis has shown great potential in public opinion monitoring, market analysis and other fields. However, traditional methods have limitations in dealing with complex semantic and emotional diversity. Therefore, this study proposes a knowledge-enhanced pre-trained language model for text sentiment analysis. The model effectively improves the model's ability to understand emotional semantics by integrating external knowledge bases, such as emotional dictionaries and domain-specific knowledge. In the pre-training stage, we adopted a large-scale Chinese text dataset and combined emotional labels for joint training. The experimental results show that compared with the baseline model, the accuracy of the proposed model in the sentiment classification task is improved by 8.3%, and the F1 score is improved by 7.5%. The model performed equally well in the fine-grained sentiment analysis task, with accuracy and F1 scores improving by 6.2% and 5.8%, respectively. In addition, the model shows stronger robustness when dealing with long texts and complex emotional expressions. Further analysis shows that the knowledge enhancement module effectively improves the model's ability to recognize emotional vocabulary and tendencies. This study provides a new technical path for text sentiment analysis and a valuable exploration for applying pre-trained language models in specific fields.