Improvement on Automated Thai Assignment Scoring by Using a Thesaurus
Keywords:automated Thai assignment scoring, thesaurus, document clustering, document classification, machine learning
Essays are a great useful tool to assess students’ learning outcomes. Automated essay scoring represents a practical solution to the time-consuming activity of manual scoring of students’ essays. However, one problem of the efficiency automated essay scoring is the synonym. The objective of this research was to compare the results of an experimental study of automated Thai assignment scoring with a thesaurus to solve the synonym problem and without a thesaurus. The experiment was conducted with 1,000 undergraduate students who were assigned to complete a particular on-line exercise in the course of Information Technology for Learning at the Thepsatri Rajabhat University, Lopburi, Thailand. The proposed model has 2 types of categories, Thai Assignment Scoring without Thesaurus (TASWOT) and Thai Assignment Scoring with Thesaurus (TASWT). Both types were designed using the principles of text mining including Document Clustering and Document Classification. The accuracy rate of classification was used to evaluate efficiency. The experimental results showed that the performance of the automated Thai assignment scoring with a thesaurus equaled 83.77 percent of average accuracy, while the automated Thai assignment scoring without a thesaurus had 74.92 percent of average accuracy. The results of the experiment indicated that automated Thai assignment scoring with a thesaurus had higher efficiency than automated Thai assignment scoring without a thesaurus. Because, that the use of Thesaurus to solve the synonym problems before clustering has resulted in some documents shifted into other clusters when comparing with clustering without a Thesaurus. As a result, machine scoring has been changed and that also leads to the effectiveness of machine scoring getting better.