A Machine Learning Based Classification of Students' Algebraic Responses Using MathBERT Embeddings

Conduction of assessments and analysing the students' knowledge is the most important part of any teaching learning process. Traditional manual grading of written mathematical solutions can be time-consuming and laborious, especially for a larger group of students. It takes lot of time and effort to correct these answer scripts. In order to evaluate students' open responses for algebraic inquiries automatically, this project explores the use of MathBERT embeddings, an advanced natural language processing technique with different machine learning classifiers. The model tries to classify the student's responses to either correct or wrong. XGBoost Classifier on the MathBERT embeddings exhibited the best accuracy for this model with 75.3% mean accuracy with standard deviation of 0.016.