Short descriptive answer evaluation using word-embedding techniques

The score given for short answers may vary from instructor to instructor. There are many short answer grading and essay grading systems existing that are either automated or semi-automated. Automated grading systems reduce human effort, but no popular tool still exists. In this paper, we are focusing on short-answer grading systems. We use simple and effective methods to evaluate short descriptive student answers. The similarity between each student's answer with its model answer is evaluated using word embedding algorithms. The similarity score is used to calculate the score. The accuracy of the scores obtained in the case of each algorithm is calculated and analyzed method-wise.