Web based Automated Essay Grading System using Latent Semantic Analysis Method for Indonesian Language

A.A.P. Ratna, N.A. Gunawan, Adhe W, N. Evianti, and B. Budiardjo (Indonesia)


) in Singular Value Decomposition divided into 3 (three) matrix. In reconstructed matrix from decomposed matrix using SVD will be seen strong correlation between topics or sentences joined in a specific group.


: In this research, we are using the LSA technique for Indonesian Language (Bahasa) and prove that we achieve 83 % for the agreement with Human Raters. Grading system is a mechanism used to determine student’s ability of the given material of the studying process. Essay is one form of grading, where there are no choices of answer, and student must answer in sentence. Essay answers may vary greatly between each exam participant, depending on their own thought. 2. Automated Essay Grading with LSA Essay type is a form of evaluation where the option of answer is not provided, and student must answer in sentence, thus the answers may vary depending on each student’s thought. Methods used in automated essay grading system nowadays are still under research because they need to follow some specific rules. Some of the methods are Natural Language Processing, Statistical Approach, Bayesian Text Classification, and Latent Semantic Analysis [1]. One of the grading methods, i.e. Latent Semantic Analysis (LSA), which uses matrix algebra to compare between expected answer and student’s essay answer [2]. Essay grading remains as an option for teacher to evaluate student’s ability, even though it is not easy to give objective judgement to every student in essay grading. Essay considered as a powerful tool to achieve studying result by many researcher, also to evaluate the thinking ability in high level such as in synthesis and analysis. [4] One of the techniques is Latent Semantic Analysis (LSA) method, it is a method to extract and represent sentence with mathematical or statistical calculation from large amount of texts [1]. The mathematical calculation is done by mapping, whether there is or there is not a word from group of words in matrix. The strength of the LSA technique lies in the syntax structure insensitive, thus the words processed are words from a bag of words ignoring the sequence of the sentences that use the words [7]. In this paper, we are using the LSA technique for Web Based Automated Essay Grading for Indonesian Language (Bahasa) and prove that we achieve 83 % for the agreement with Human Raters. Key Word: On-line examination, LSA, agreement with human raters, automated essay grading system, singular value decomposition. This method defines the similarity of word or passages in large group of texts. It is based on Singular Value Decomposition (SVD) which is as mathematical matrix decomposition technique. Matrix formed from whether there is or there is not specific words (usually already defined as

Important Links:

Go Back