A decision support system for evaluation of essay type questions

In this thesis study, a decision support system was developed for instructors to aid them during evaluation of short answer/essay type questions. The aim of this study was to devş6elop a system that will help the instructor to make a fair grading by comparing his/her grade with the system's suggested grade. Zemberek NLP Library was used for separating given answer (by the student) and the correct answer sentences ş2to words and words to roots. This library was also used for spell checking and correction of the typos in the given answers. Vector space model and cosines similarity theory were used to calculate the similarity between correct answers and the given answers. The scientific significance of the results of developed algorithm was tested by calculating the Pearson's correlation and by applying t-test between actual grades (given by the instructor) and the grades calculated by the system. The Pearson's correlation coefficient between given grades and the system grades was found to be 0,667. T-test analysis results have also given very similar results to Pearson's correlation coefficient. These results indicated a moderately strong relationship between the given grades and the system grades.