Graph based Hybrid Assessment System for Programming Assignments

Abstract

Programming is a practical process; students need to write a lot of programs in order to master it. However, with large number of students, the assessment of programming exercises leads to extensive workload for teachers making it difficult for instructors to provide constructive and corrective feedback or even additional help when the students need it. In this work, we address the issue of automatic assessment for programming assignments. The goal of which is to provide immediate grading and comprehensible feedback to the learners, while taking some of the workload burden off the teachers. This paper proposes a system combining results from dynamic and static analysis to ensure a reliable and objective evaluation job. While dynamic analysis is based on unit testing framework, the static analysis will quantify the structural similarity between students' programs and the solutions provided by the teacher. In order to perform such comparison, a suitable program representation and an adequate similarity measure will be presented.

Validity of a graph-based automatic assessment system for programming assignments: human versus automatic grading

“…In fact, students' programs are run through a predefined set of data, and afterwards their outputs are compared to the expected answers. The approach is described in more detail in a previous paper where it was suggested the use of XUnit, a dedicated framework to automate and conduct tests in a given language [31]. XUnit is additionally used to supply significant and comprehensible feedback.…”

“…This measure is based on iterative calculation of similarity and the principle that two nodes are as similar as their neighbors are [32]. The calculation of similarity is detailed in a previous work [31] and given by (1). where 𝑥 𝑖𝑗 is the calculation of nodes i and j similarity in (k+1) iterations.…”

“…Both the similarity of nodes and the similarity of CFGs take values in the [0, 1] interval. More details were given in a previous paper [31].…”