Latent Semantic Analysis and Winnowing Algorithm Based Automatic Japanese Short Essay Answer Grading System Comparative Performance

In this paper, advanced of research on e-learning application for short essay grading system had been conducted. This system was developed based on the needs of Japanese Language study program for short essay examination that required time and focus for finishing those tasks. Human abilities are limited by their energy, so that cognitive assessment objectivity could decrease in line with the elapsed time. Latent Semantic Analysis (LSA) and Winnowing Algorithm are two methods used in developing the automatic short essay answer grading system called SIMPLE-O by Department of Electrical Engineering, Universitas Indonesia. These two algorithms are chosen based on its ability to do semantic analytic without the needs of understanding about the characteristic of its languages. LSA used Singular Value Decomposition (SVD) as its main method, besides Winnowing algorithm is based on fingerprinting. These algorithms are applied into the automatic system to assess the Japanese language exam with close results between them with average accuracy of Winnowing algorithm is only 1.06% lower than LSA that could gain 87.78%. These two algorithms should be suitable for grading short essay answer in Japanese language.