Mathematic Problems Categorization Using Synthetic- LLM Models

Accurate identification and understanding of mathematical problems are essential for effective problem-solving. This study addresses challenges in classifying and tagging mathematical questions for educational purposes, such as ambiguous phrasing, varied structures, and the complexity of different mathematical concepts. To tackle these issues, the proposed method consists of two main steps: utilizing the BERT language model for classifying questions related to derivatives and integrals, and employing the T5 model for precise tagging of problems. The process involves tokenization and thorough training of both models on a custom dataset that includes four categories of mathematical topics, focusing on integral and derivative problems. Evaluation is conducted using five metrics: accuracy, precision, recall, F1-score, and confusion matrix. Results indicate a high level of accuracy in classifying question types and tagging, highlighting the effectiveness of the proposed approach in enhancing educational tools for mathematics through advanced language processing techniques.