Contextual Emotion Detection of E-Learners for Recommendation System

In the recent pandemic times, there was an impactful transformation in imparting education which required everyone to become online learners. There has been an exponential growth in the number of e-learners attending classes online and taking MOOCs courses. This has opened an avenue for research to analyze the emotion of e-learners through reviews of students to evaluate the learning outcomes and performance of the course. Most challenging task is to find the exact pulse of the e-learners’ emotions from the huge data of the e-learners reviews. The reviews on all online platforms are mostly textual and this qualitative data needs to be quantified for analysis. There is a necessity to propose contextual emotion detection of e-learners by extracting the relevant information which can be correlated to the performance of the course on e-learning platform. Further, it can be a recommendation system to the aspiring e-learners to make decision based on the satisfaction index of previous e-learners. This paper leverages deep learning techniques to train various models for academic emotion detection using dataset E-Learners Academic Reviews (ELAR) prepared from online textual feedback of e-learners and MOOCs course reviews. The Bidirectional Encoder Representations from Transformers (BERT) transfer learning model used to detect the emotions outperformed the other models. This proposed method using ELAR dataset is a novel approach to identify the right emotion of e-learners from the course reviews available on e-learning platform. The results were discussed with a benchmark of ISEAR (International Survey on Emotion Antecedents and Reactions) and GoEmotion Dataset.