<![CDATA[Essays play a crucial role in traditional assessments, but evaluating them accurately, efficiently, and fairly poses a major challenge for educators. Automated Essay Scoring (AES) aims to address this issue by leveraging computational techniques to support teachers in the grading process. This study explores a classification model to classify the score based on the feature we created. We incorporate additional features aligned with the ASAP 2.0 scoring rubric, such as Lexical Sophistication, Source Adherence, Novelty and Relevance, and a Semantic Disruption feature. These features are used to construct a distributed representation of essays, which is then input into a Support Vector Machine (SVM) model for holistic score prediction. The proposed model achieved a Quadratic Weighted Kappa (QWK) score of 0.8397, indicating a high level of agreement with human raters. The results demonstrate the effectiveness of combining rubric-informed features with a non-linear classifier. The findings can be implemented for educational settings, where the model can be utilized to provide scalable and consistent scoring support, reduce grading workload for instructors, and deliver timely feedback to students. By aligning with rubric-based criteria, the approach can also foster more transparent and constructive learning processes, helping students identify specific areas for improvement in their writing. While the model exhibits strong predictive performance, it also presents limitations related to interpretability and generalizability, especially across diverse writing prompts and domains.]]>
Copyrights © 2025
