Text Classification and Similarity Algorithms in Essay Grading

Open and closed questions are both important in an examination to evaluate the teaching and learning process. Both types of questions have their own advantage and disadvantage. Open-ended questions in the form of essays require manual grading that is harder and time-consuming. This paper presents an effort to grade essay answers based on classification techniques and similarity algorithms. We examine 1648 essay answers from an examination of Bahasa Indonesia. Teachers use integer values in the range of 0–12 to score the answer. Our result suggests that both classification techniques and similarity algorithms result in about the same grading as shown by the RMSE values in the range of 2.7–3.0. Hence, the use of either technique is apparently acceptable depending on the situation.