Maksimov Denis Romanovich (National Research Nuclear University "MEPhI")
Chervakov Philipp Sergeevich (National Research Nuclear University "MEPhI")
| |
The article presents a method for the automated generation of analytical reports on student academic performance using a fine-tuned GPT-2 language model. The relevance of this study is driven by the growing volumes of educational data and the need for timely analysis of student performance to provide prompt feedback. The authors compiled a specialized dataset containing students' academic indicators and instructors' textual reports. Using this corpus, GPT-2 (774M parameters) was fine-tuned leveraging the HuggingFace Transformers library and PyTorch. As a result, the model effectively generated human-like analytical texts. Evaluation included automatic metrics (BLEU, ROUGE) and expert assessments by instructors. Generated reports demonstrated high accuracy (4.8/5), clarity (4.6/5), and usefulness of recommendations (4.7/5). Automation significantly reduced the workload of educators and enhanced personalized feedback accessibility for students. Discussed limitations include the necessity for high-quality training data and the model's limited context window. Future research directions encompass employing more advanced language models and integrating generated reports into educational practices to evaluate their impact on learning outcomes.
Keywords:analytical reports, student performance, automatic text generation, natural language processing, GPT-2, neural networks, machine learning, model fine-tuning, educational data, feedback.
|
|
| |
|
Read the full article …
|
Citation link:
Maksimov D. R., Chervakov P. S. AUTOMATIC GENERATION OF ANALYTICAL REPORTS ON STUDENT PERFORMANCE BASED ON THE ADVANCED GPT-2 LANGUAGE MODEL // Современная наука: актуальные проблемы теории и практики. Серия: Естественные и Технические Науки. -2025. -№06/2. -С. 135-140 DOI 10.37882/2223-2966.2025.06-2.21 |
|
|
LEGAL
INFORMATION:
