Sentiment Analysis of Chinese Online Course Reviews by Combining Prompt Learning and Contrastive Learning
Kai Xu, Junfei He · 2023
The study of sentiment analysis of Chinese online course reviews can provide an in-depth understanding of students' learning attitudes and needs, optimize the teaching mode, and thereby improve the quality of teaching. To this end, this research proposes a sentiment analysis method for Chinese online course reviews, which combines prompt learning and contrast learning. First, a question and answer prompt template for the large-scale language model ELECTRA is constructed to guide ELECTRA to learn the semantic information contained in the Chinese online course review text. Then, contrastive learning is used to further enhance the data samples and learn feature representations. Finally, the feature representation trained by contrastive learning is input into the sparse Softmax function to perform sentiment polarity classification of Chinese online course reviews. The experimental results on real Chinese online course data sets show that the macro F1 value of the proposed model reaches more than 85% on the three data sets.