Research on Sentiment Analysis of Online Course Evaluation Based on EN- BERT-CNN
Yuan Guo, Hongmei Li · 2024
At present, BERT model has achieved great success in text emotion analysis, and has achieved excellent performance on many benchmark data sets. Research shows that by fine-tuning BERT model, its performance in text emotion analysis tasks can be further improved. In this paper, EN-BERT-CNN model is proposed, that is, the first token (CLS) vector of each encode layer of BERT model is regarded as a sentence vector, and then CNN convolution neural network is used to extract features of different levels for emotion classification. This method integrates the information of the bottom layer network, middle layer network and high layer network of the model better, realizes the combination of word vector and sentence vector, and improves the accuracy of data processing by the model. Through the comparison of experimental data, it is proved that the accuracy of model data analysis has been improved, and the accuracy of massive open online course course review analysis has been further improved, which also provides a new idea for better solving the task of emotion classification.