Analysis of Online Education System of Bangladesh during COVID-19 Pandemic Based on NLP and Machine Learning: Problem and Prospect

Abstract

The Coronavirus Disease (COVID-19) pandemic has interrupted the education system throughout the world. Bangladesh is no unlike; all educational institutions are shut down across the country. The online teaching method is quite new especially for the developing countries like Bangladesh. Therefore, the main aim of this work is to mine student’s opinions about online class during this COVID-19 pandemic. To achieve this aim, this paper uses a questionnaire survey through the google form to collect Bangladeshi student’s opinion on online class, build a corpus of 5005 data containing both Bangla and Romanized Bangla text. After data pre-processing and extracting the features, machine learning classifiers were deployed. Then performance measurement was done in terms of accuracy, precision, recall and F1 score. In the final evaluation, we achieved highest of 80% accuracy with SVM classifier, where the accuracy achieved by Logistic Regression, Random Forest and Multinomial Naïve Bayes classifier was 78%, 77% and 77% respectively. We tried to predictthe problems faced by students and suggested possible solutions about online class. The result showed that 27.9% student faced financial problem and 25.8% student faced unstable internet problem. 54.8% user suggested stable internet facility in low cost or free and 23.1% suggested financial assistance for online class as the possible solution of aforementioned problems.