Abstract:

Abstract In the present age of advancement in computing through the application of artificial intelligence, a host of programming and modules are designed to facilitate a virtual learning environment, each claiming its own efficacies and usefulness in virtual learning during the pandemic. The present paper endeavors to design a unique model, named for the first time as EdBERT, for sentiment analyses of virtual learners with most accuracy of their review classification. The model focuses on an improved version of sentiment analyses with Google BERT while facilitating educational feedback corpus. The methodology used is a comparative study using the tool ‘fine-tuned Google BERT’, which is trained at three different stages for understanding the language, context, sentiments and thus, performs classification of learners’ feedback accurately. The model and its functioning are given in the discussion with valid proofs of accuracy testing and analyses. EdBERT stands as a state-of-the-art model in AIEd sentiment analyses with the best evaluation matrix so far with 87.89% accuracy, 88 % F1- score, 89 % Precision, 88 % Recall values. These values are of evaluation matrix is better than any other recent models discussed in the article. AIEd is comparatively new domain which is getting explored by academic researchers and scientists to im-prove the productivity of the learners, instructors and the learning environment. This work is a deep learning and natural language processing models can be used to provide reliable sentiment analysis with three basic sentiments class. Further this work can be extended with Plutchik’s wheel of emotion that will help in capturing the emotions with help of AI and Deep Learning more correctly.

Keywords: