A deep learning model for automated marking of students’ assessments in a learning management system (LMS)

Learning Management Systems (LMSs) are widely used to support teaching and learning, with platforms such as Blackboard managing lectures, activities, assessments, and reports. Although LMSs provide useful tools and some automated feedback, the accuracy of evaluating students’ typed responses has received little attention in prior research. A particular issue arises in fill-in-the-gap questions, where answers are marked only if they exactly match the instructor’s input, often leading to unfair grading for minor spelling errors. To address this problem, we propose a model that integrates the Levenshtein edit distance with deep learning methods to identify and correct spelling errors, enabling fairer and more accurate automatic grading. The model demonstrated strong performance, achieving an average F1-measure of 0.938 on a dataset of misspelled words.