Enhancing Large Language Models for Precise Classification of Teacher Praise Discourse: A Fine-Tuning Approach

In the field of education, teacher praise, as a form of positive feedback, is regarded as a crucial mechanism for fostering student development and maintaining a positive learning environment. In recent years, large language models, particularly those represented by the GPT series, have achieved breakthroughs across multiple dimensions of natural language processing, demonstrating exceptional capabilities in general language understanding and generation tasks. However, when focusing on fine-grained analysis in specific domains, such as the precise classification of teacher praise discourse, large language models without specialized optimization often fail to achieve optimal performance. To address this, our study proposes a strategy aimed at enhancing the performance of large language models in the teacher praise classification task through fine-tuning techniques. We first established a dataset of teacher praise discourse. Using this dataset, we then conducted targeted training on selected large language models to improve their accuracy and robustness in the teacher praise discourse classification task. Experimental results show the fine-tuned model outperforms baseline models on datasets.