A Change Detection Method of Arguing Points by Factor Analysis for Case Method Support

We are developing the facilitation system for learners´ discussion in the case method that allows the learners to acquire the knowledge of problem-solving with exchanging their ideas on the practical case. The facilitation system captures the learners´ speeches in real time, detects the change of the main arguing point and gives the advises to the learners. However, the changes of the arguing points can not be detected by a basic method. In this paper, we propose a change detection method of the arguing points based on the frequency of the related words to each arguing point in each speech. Because the related words to several arguing points are included in the speech, the frequency of words that are not related to the main arguing point should be ignored. If the arguing points are related to each other, the words that are not related to the main arguing point often appear in the speech. When the correlation between the frequencies of the arguing points in the last n speeches is judged to be high by factor analysis, the frequencies are ignored for identifying the main arguing point. As a result of an experiment on two kinds of the case method, it has been confirmed that the recall rate and the precision rate of the change detection are improved by 16% and 7%, respectively.