Graph-Based Semantic Similarity for Programming Exercise Assessment: A Survey

In the present knowledge proliferating from computers, it has become a necessity to learn computer programming languages. Competent and proficient programming skills are very important for any computer science student, and to obtain proficiency, learning these skills requires much practice using programming exercises. However, as the sizes of classes continue to increase, assessing the programming exercises manually creates an overwhelming workload for instructors, adding significant amounts of time and culminating in errors in grading. Current tools typically provide feedback by listing failed test cases, but this can be tedious and incomplete. This paper describes a design for an automated assessment tool that evaluates the accuracy of programming assignments, provides grades and counterexamples, and compares with a reference solution. Rather than count the number of tests that were completed, our tool supplies a semantically different execution path between a student’s submission and the reference solution. Index Terms-programming skills, automated assessment, automated grading.