Engagement measurement of a learner during e-learning: A deep learning architecture

Teaching in online mode is challenging, but measurement of student engagement is even more challenging and complex but crucial to overall learning of the participants. Although online environment improves the overall learning and removes the barriers of time schedules, it requires a high level of self motivation from the learners. Those at the brink of self motivation may perform exceedingly better under online learning if a tool can help them concentrate. These tools require the overall engagement level of the learner. The current manuscript proposes a multi module deep learning based engagement level predictor model. Basically, three modules, a convolutional neural network (CNN) module-for hierarchical feature extraction and other two modules-LSTM and attention based network, for temporal modeling of task are utilized in the proposed architecture. A dataset consisting of 9068 video sequences is used to measure the engagement level of the learners. The proposed model also introduced a novel cost function for prediction. Experimental results validate the usefulness of proposed loss function over traditional loss functions.