Artificial Intelligence-Enabled Essay Grading System Using NLP Via Semantic Similarity Analysis and Supervised Machine Learning Techniques

• Jide E. T. AKINSOLA

  Author

• John A. OLADITI

  Author

• Ifeoluwa M. OLANIYI

  Author

• Emmanuel A. OLAJUBU

  Author

• Ayomide O. EMMANUEL

  Author

• Ganiyu A. ADEROUNMU

  Author

A manual grading system can be affected by several factors such as subjective judgment of the reviewer, tiredness, sensitivity to handwriting, as well as the reviewer’s emotions, physical and mental state, which may lead to unfairness and inconsistency in student grades. In order to ensure fairness in computing the grades of students, it is good to embrace the use of an automated grading system which ensures consistent and objective grading. This study employed a numerical representation of semantic similarity by comparing the student's answer with both the examiner's answer and the comprehension passage, which was later combined using weights of 0.05 and 0.95. Random Forest (RF), Decision Tree (DT), and Gradient Boosting Regressor algorithms were trained and evaluated using R-squared value, MAE, MSE, and RMSE. The performance evaluation results show that the Decision Tree model gave the best results, achieving the highest R-squared score of 0.9717 and the lowest prediction errors of 0.1030 MAE, 0.0703 MSE, and 0.2651 RMSE, which suggests that the combination of semantic similarity and question score provides useful information for estimating student performance. However, similarity alone may not fully capture the overall quality and correctness of an answer. The study, therefore, suggests that future improvements could incorporate features such as grammatical correctness, relevance, completeness, and factual accuracy.

• 44_Akinsola et al. (2026)

Copyright (c) 2026 Jide E. T. AKINSOLA, John A. OLADITI, Ifeoluwa M. OLANIYI, Emmanuel A. OLAJUBU, Ayomide O. EMMANUEL, Ganiyu A. ADEROUNMU (Author)

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.