Machine Learning Algorithms for Improved Script Evaluation

Evaluating subjective answers in educational assessments presents a complex challenge due to the need for contextual and semantic understanding. Traditional keyword-based approaches often fall short in capturing the subtleties of students' responses. This paper investigates the use of three different methods: BERT (Bidirectional Encoder Representations from Transformers), Cosine Similarity, and Fuzzy Logic to enhance the accuracy and context sensitivity in evaluating subjective answers. The experimental analysis demonstrates that these methods can improve the evaluation process by offering more precise and contextually aware grading. A comprehensive comparison was conducted using metrics such as accuracy, precision, recall, which resulted in improvements, particularly with BERT. The implications of incorporating these methods in educational settings are profound, offering potential advancements in automated grading systems. Overall, key findings suggest that integrating these methods can lead to more reliable and efficient subjective answer evaluation, contributing to the advancement of automated educational technologies.