Automatic Essay Scoring for E-Learn System

Authors

  • Ameer Hassan Hadi, Ahmed Hussein Aliwy

DOI:

https://doi.org/10.17762/msea.v71i3.264

Abstract

The e-learning system is used to support and enhance the educational process with many facilities over traditional learning. Some of these facilities are electronic exams and automatic scoring for answers of types true and false, multiple choices, and may be short answers. Therefore, many researchers take this research direction but the biggest challenge of exam scoring is the scoring of essay exams, which is an open problem. Automated essay scoring is an educational evaluation technique and part of the natural language processing (NLP) application. Numerous considerations, including cost, accountability, standards, and technology, contribute to the increased interest in automated essay scoring. In this work, a complete essay scoring system is proposed with different methodologies that are evaluated with different evaluation metrics. These methodologies are used for classification\regression tasks; (i) each one of nine classifiers/regressions is used as an independent classifier/regression; (ii) the best three classifiers/regressions algorithms are used for getting the score as the average, and (iii) using augmented combination of all the classifiers\regression for calculating the final score. Four categories of features are used; raw features, morphological features, compound features, and orthographical features with weight for each feature that reflect the feature importance. The results, on the Hewlett essay scoring dataset, showed that scoring using the average of the best three classifiers and augmented combination have the lowest error rate in most tests. Also, they are more stable than the other classifiers where there is not any huge rising in errors in all the tests.

Downloads

Published

2022-06-09

How to Cite

Ameer Hassan Hadi, Ahmed Hussein Aliwy. (2022). Automatic Essay Scoring for E-Learn System. Mathematical Statistician and Engineering Applications, 71(3), 1011 –. https://doi.org/10.17762/msea.v71i3.264

Issue

Section

Articles