Automatic Short Answer Scoring (ASAS) Using String-based Similarity and Query Expansion

Abstract

Advances in technology and information in the field of education have made the learning system computerized, where the learning process and assessment process can be carried out using electronic media connected to the internet. The assessment process in elearning is used to measure students' ability to understand the learning material obtained during the learning process. In e-learning, the assessment process for essay exams is still done manually by educators, which requires a lot of time and energy, so there is a need for an automatic essay exam assessment system. This research aims to assess the short answer essay exam using the string-based similarity method combined with the query expansion method with stages carried out including case folding, tokenizing, stop word, stemming and calculating similarity scores using the cosine and Jaccard methods and their modifications as well as understanding the meaning. Synonyms of the words being compared using the query expansion method. The results of the correlation test showed that the weighted cosine coefficient and weighted Jaccard coefficient methods combined with query expansion have the highest correlation value of 0.913 to 0.943, however, the comparison test shows that the weighted cosine coefficient method combined with query expansion produces a decision that there is no difference in average value - The average between the results of the educator's assessment and the results of the automatic short answer scoring program assessment using the weighted cosine coefficient method and the query expansion method.