Automated Essay Scoring: A Comparative Study of Machine Learning and Deep Learning Approaches

Automated Essay Scoring (AES) refers to the use of technology to automatically assign scores to essays. It serves as a viable alternative to traditional human grading methods and has become an important tool in high-stakes educational assessments. It has received considerable research attention in academic institutions as it accelerates the process and alleviates the workload on human evaluators. Hence, we studied the Artificial Intelligence and Machine Learning methods that were employed to develop automatic essay scoring systems. We have discussed the limitations in the existing research and explored emerging trends in this area. This study seeks to compare the performance of several common AES frameworks, emphasizing on machine learning with complex language features and deep neural algorithms. Through rigorous testing, the strengths and weaknesses of both traditional and modern models were assessed under varying circumstances and the results were evaluated using the efficiency score which were compared to the grades given by human raters. The results revealed that the neural network model, trained on comprehensive prompt data and employing Natural Language Processing technique for feature extraction, outperformed all the other models. This indicates that it gave a closer alignment with results of the human raters as compared to the conventional Machine Learning models.