Similarity Detection of Student Assignments Using Rocchio Method

Abstract

Student assignments is one component that influences the final score of a course given by the lecturer. The assignment of a course is charged at 20% of the final assessment of a course. Each lecturer has an average of 40 students in each class, while the lecturer can teach more than one class of courses. The large number of students will make the number of assignments too large because the ratio of the number of lecturers and students is very different. This is a burden on the lecturer because it has the responsibility to check each of student work assignments. The level of similarity of course assignments is one of the ways lecturers can assess a subject's assignment. This study aims to help lecturers detect the similarity of student tasks by using a task similarity detection system. This study uses Rocchio method to detect the similarity of words. The type of document tested is formatted in .pdf. The language used in the course is Indonesian. The preprocessing stages in this research consists of Case Folding, Tokenizing, Filtering, Sorting. The TF.IDF uses as the word weighting and the similarity measurement with the Rocchio method so that it gets the percentage value of the similarity between documents. The development stage started from planning, analysis, design, implementation and testing. The testing stage including testing with dummy data and usability data. Based on the research that has been done, the Rocchio method can classify documents that are similar to the results of the accuracy of 74,6% and detection errors from the similarity of queries with documents of 2.54 or 25.4%.