Natural Language Processing Research

In this paper, a web-based Turkish automatic short answer grading system (TASAG) is developed to score exam questions and to generate online exams for automatic scoring of short-answer questions. The novelty of this study is that TASAG is the first software of its kind for the Turkish language. The algorithm of the TASAG software is a hybrid that determines which method will be used at runtime based on the word number dimensions to achieve accurate scoring. To measure the software's accuracy, a case study is performed for computer engineering. Instructors' scoring results and the TASAG software scoring results were compared. Two instructors prepared different answer keys for the same exam to increase the accuracy of the scoring. The scoring results, which are compared in the figures, are very close to each other, which indicate the effectiveness of the TASAG software. Moreover, the TASAG arithmetic mean scores for each answer key are calculated and given as the final score for the exam. According to the results, the TASAG software can be used for automated short answer grading system in Turkish with a 92% success rate.