Performance evaluation of an intelligent multimedia learning assistant platform

Recently, the literature reported that multimedia learning is more appealing and can increase learning effectiveness. Furthermore, cued retrospective reports (CRRs) can enhance students´ learning performance, especially for the novices lacking prior knowledge. In this work, a multimedia learning assistant system is constructed by applying an eye-tracking technique to record and evaluate learner´s CRRs. First, each student takes the test after watching the multimedia teaching animations. Those who cannot pass the test will be provided to watch the same video but with segmented video clips to reduce possible high cognitive load. Notably, one-class SVM classifier was applied to automatically judge whether the learners´ concepts are correct. Experts´ or high achievement peers´ CRRs are provided to assist the low-achievement learners. Then, those low-achievement learners accept the second test to verify the effectiveness of our proposed system. 31 students in a junior high school participated in the pre-test, and 16 students did not pass the first test after watching the video. The experimental results revealed that these 16 students have increased their grades from 53.75 to 75. In addition, the classification rate can reach up to 94.59%. The proposed learning assistant system can effectively assist teachers in providing realtime assessment and feedback.