This study explores the application of sentiment analysis within the framework of intelligent governance to enhance the responsiveness and quality of public services. Leveraging recent advances in natural language processing (NLP), we develop a hybrid Bidirectional Encoder Representations from Transformers (BERT) model for fine-grained sentiment analysis of public opinion texts. The proposed approach integrates BERT’s contextual embeddings with the Transformer’s multi-head attention mechanism to improve sentiment feature extraction and classification accuracy. Experimental evaluations on multiple public and proprietary datasets demonstrate that our model achieves an accuracy of 0.913, outperforming standard BERT and other deep learning baselines (by up to 12.8% on application datasets). These results substantiate the model’s effectiveness in capturing subtle variations in sentiment polarity. The findings highlight the potential of advanced NLP techniques to support data-driven decision-making in intelligent administration, facilitating more adaptive and citizen-centered governance.
Jiang et al. (Fri,) studied this question.