Peer Assessment Based on the User Preference Matrix

Abstract

Peer assessment not only provides a solution to inefficient teacher-student interaction between the students and teachers at universities, but can effectively improve the students’ learning efficiency. However, students are seldom engaged in peer assessment tasks and the scores they get from peer assessment differ from those given by the teachers. An automatic ranking method for English essays based on the user preference matrix and the stochastic gradient descent method was proposed in this study. Experiments showed that this new method could improve the performance of the stochastic gradient descent method when the credibility of the student’s evaluation was considered. Teachers can obtain the rankings of students’ essays by analyzing a few peer-assessment scores obtained by this method and optimize their teaching strategies accordingly.