An Attempt to Detect Learner’s Hesitation by Machine Learning in Word-Reordering Problems

Abstract

　It has lately become easier to store learners’ study logs in educational settings, with the advancement of e-Learning systems. Mouse trajectory data is also such an example, by interpolating discrete X and Y-coordinates. Our study group has been involved in the development of a Web application to grasp when learners hesitate in the process of solving English word-reordering problems. The information of learners’ hesitation would be an important clue to know learners’ understanding. Our system provides an interface which records and reproduces learners’ mouse trajectories with click logs for retrieval and analysis of the data. In this study, we adopted a supervised learning technique in machine learning (using random forest algorithm) to detect above-mentioned hesitation and achieved about 82% precision of estimation. Furthermore, this article dealt with the investigation on the use of the “grouping” function for multiple words to be moved together, and the result showed the possibility that this function was used positively to construct phrases and idioms.