Quality control method for peer assessment system based on multi-dimensional information

Abstract

In recent years, online learning has received widespread attention, and has played an important role in recent years. As a new application field of crowdsourcing system, peer assessment can solve student’s performance evaluation problem in massive online courses. Traditional crowdsourcing quality control algorithm has not been able to use effective information to the evaluation of workers. Aiming at this problem, a quality control algorithm based on multi-dimensional information is proposed. The user’s behavior, comment text information and other useful elements are combined together. The feature vectors of reliability are extracted from a variety of information based on the frame of log-linear model. And then, the gradient descent algorithm model is used to study the optimal parameters. The experimental results show that when comparing with the traditional Expectation Maximization algorithm, our multi-dimensional quality control algorithm has better performance in the accuracy and mean square error.