Prototype of Automatic Essay Assessment and Plagiarism Detection on Mobile Learning "Molearn" Application Using GLSA Method

Abstract

In evaluating the student’s learning outcomes, essay exams were commonly used by teachers to measure the level of student’s understanding of the learning material. However assessing essay answers was more difficult in reality because it contained teacher’s subjectivity and required a longer correction time. In addition, detecting similarity in essay answers between students also required more teacher’s efforts. In previous studies, a prototype of essay answer assessment and plagiarism detection had been successfully created. However, the prototype display still needed an improvement based on the evaluation results given by biology teachers in East Java Province as the application users. The previous prototype also still carried the Latent Semantic Analysis (LSA) method which had several weaknesses. Therefore, this study aimed to produce prototypes that had better display and text similarity methods. The Generalized Latent Semantic Analysis (GLSA) method was chosen because it was able to cover the weaknesses of the LSA method. GLSA was able to detect sentences that had syntactic errors or missing common words. Based on the evaluation results, this study succeeded in producing a prototype with a better display value. The level of user satisfaction increased by 6.12%. In addition, the study succeeded in using the GLSA method as a substitute for LSA for creating better prototype essay assessment and automatic plagiarism detection.