Sentiment Classification of Chinese Comments Based on Pre-trained Language Model

The rapid development of social media has made people's ways of expressing emotions more diverse, and sentiment analysis for Chinese texts has become a hot research topic. The application of deep learning models has dramatically improved the effectiveness of text sentiment analysis, the application of large-scale pre-trained models has made natural language processing more effective. This article designs a Chinese text sentiment analysis model based on BERT-BiLSTM-CNN. Firstly, the pre-trained model BERT is used to encode the text. Then, BiLSTM is used to extract contextual information and features from the text, and CNN is used to extract critical local features from the text. Finally, comparative experiments were conducted on the dataset, first comparing the results with deep learning models, and then comparing them with BERT based models. The experimental results show that the model designed in this paper has high accuracy in the classification task of Chinese text sentiment analysis.