Deep Sentiment Analysis: Exploring Emotional Trends in Architectural Decoration Language Using the BERT Model

With the proliferation of social media and online review platforms, the importance of sentiment analysis in the fields of architecture, design, and decoration has become increasingly prominent. While traditional machine learning methods have made strides in this regard, deep learning, especially BERT and its variants, has been demonstrated to have significant advantages in various natural language processing tasks. This study aims to delve into the application of BERT in sentiment analysis within the realms of architecture, design, and decoration, further optimizing its performance. To this end, we introduce regularization and data augmentation strategies to enhance the model's generalization capability. Through experiments on the AIDA dataset, we validate the efficacy of the proposed approach. Experimental results indicate that the BERT model combined with regularization and data augmentation outperforms other methods. Additionally, we provide a detailed error analysis, offering directions for future research. Overall, this study presents a novel perspective and effective technical solution for sentiment analysis in the fields of architecture, design, and decoration.