Automated Grading of Use Cases

Use cases (UCs) play a crucial role in software engineering courses, with students frequently using them in assignments, projects, and exams. However, as the number of students enrolling in Computer Science and Software Engineering programs continues to rise, manual grading of these models is becoming increasingly time-consuming. While automated grading tools for class diagrams exist, the automation of grading use case models has received limited attention.This paper proposes an approach for automatically grading use cases. To compare a student’s solution to the teacher’s solution our approach uses structural matching, syntactic and semantic word matching, natural language processing for sentence matching, and flattening of use case hierarchies. The grading algorithm is evaluated on three actual undergraduate and graduate assignments that involve modeling a Gas Station fill-up scenario, a Supermarket checkout scenario, as well as the interactions involved in playing the board game Elfenroads. The results show that with some tuning our automatically determined grades lie within 7% difference of the instructor’s manual grades.