To keep up with the current spread of education, there has arisen the need to have automated tools to evaluate assignments. As a part of this thesis, we have developed a technique to evaluate assignments on regular expressions (regexes). Every student is different and so is their solution, thus making it hard to have a single approach to grade it all. Hence, in addition to the existing techniques, we offer a new way of evaluating regexes. We call this the regex edit distance. The idea behind this is to find the minimal changes that we could make in a wrong answer to make its language equivalent to that of a correct answer. This approach is along the lines of the one used by Automata Tutor to grade DFAs. We also spoke to different graders and observed that they were in some sense computing the regex edit distance to assign partial credit.
Computing the regex edit distance is a PSPACE-hard problem and seems computationally intractable even for college level submissions. To deal with this intractability, we look at a simpler version of regex edit distance that can be computed for many college level submissions. We hypothesize that our version of regex edit distance is a good metric for evaluating and awarding partial credit for regexes. We ran an initial study and we observed a strong relation between the partial credit awarded and our version of regex edit distance.
Computer Science (MS)
Department, Program, or Center
Computer Science (GCCIS)
Kakkar, Himesh, "Automated Grading and Feedback of Regular Expressions" (2017). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus