A Study of Multifaceted Sentiment Analysis Based on Microblog Texts

Although ceramic items are essential for everyday living, the demands of social and economic growth are not being met by the emotional analysis study on ceramic products. In order to efficiently mine the important data on ceramic emotion, this work has combined applying the BERT model with classifiers for multiple labels on the initial dataset we initially developed to provide a novel approach for the purpose of text recognition and categorization of ceramics-related microblogging data. First off, a total of 10,154 raw samples—gathered using Python from microblogging data specifically associated with the term 'ceramics'—were used to create the first dataset with multi-label emotions, including 7564 samples. Second, the information was labeled with several labels and divided into three groups: quality, appearance, and pricing information. Thirdly, a hybrid model for evaluating ceramic emotion has been presented, which combines multi-label classifiers with Bert. The results of several trials indicate that the predictor has a 92% accuracy rate and a 0.33 loss value. The annotation rules, dataset properties, and insights into various approaches for understanding emotion and cultural legacy in ceramics are highlighted in this study.