Twitter Sentiment Analysis and Topic Modeling for Online Learning

The COVID-19 pandemic has caused entirely new transitions in the worldwide higher education sectors. Online education has gained popularity due to its ability to provide more flexible access to information and teaching at any time and from any location. Students and faculty members can connect to their institution’s online portals and engage in virtual educational activities through the use of live environments provided by electronic learning management systems. Despite the fact that modern technology actively facilitates these online sessions, students’ active involvement remains a difficulty that has been addressed in prior studies. Using the Valence Aware Dictionary for Sentiment Reasoning (VADER) model, the sentiment of roughly 220,000 tweets was calculated to identify people’s opinions toward online learning by using sentiments as positive, negative, and neutral. The topic modeling was utilized to uncover some hidden themes and to define the narrative and direction of the online learning related topics. Topic modeling also helped to detect inter-cluster similar terms and assess the flow of information from a group of similar perspectives. In this study, one text feature extraction TF-IDF with LDA model was utilized for topic modeling and an optimum number of topics were extracted from the corpus that are mostly addressed in the tweets. Finally, a cutting-edge deep learning and machine learning classification model was applied to the dataset with varying epoch sizes to predict and classify the emotions of the participants and proposed a new machine learning approach to evaluate the dataset related to online learning. Multiple parameters, such as accuracy, validation loss, validation accuracy, etc., have been used to evaluate the deep learning model.