Text classification and sentiment analysis using deep learning techniques with special reference to COVID-19

Abstract

The COVID-19 pandemic has pushed the world into a highly dangerous situation, killing thousands of lives and putting the global health infrastructure at significant risk. The virus has spread worldwide, and every nation has worked together to combat against this virus. To prevent the spread of the Coronavirus. fight with them and for public awareness. Natural Language Processing approaches such as automated systems that can classify mediical documents/research articles, measure public sentiment, and detect fake news/ruimors play vital roles. This research makes significant contributions to text classification, sentiment analysis, and rumor/fake news identification on COVID-19 by applying Deep Learning (DL) and Natural Language Processing (NLP) techniques. The first contribution is a DL technique for classifying and segregating medical docunnents and research papers on COVID-19 to decrease the searching effort of the researchers for relevant content or information. The second contribution is a rumor/fake news classification model for detecting false and misleading narratives on COVID-19: from social media and online news blog to stop the spread of rumours. The third contribution is a sentiment analysis task, which has grown in popularity as an emerging area in NLP. An Aspect-Based Sentiment Analysis (ABSA) system has been propos,ed by applying Best Worst Method (BWM) of Multi-Criteria Decision Making (MCIDM) technique and Deep Neural Network to predict the polarity of public sentime:nt underlies several aspects from 1witter throughout lockdown and unlock stages iin India. The fourth and final contribution of the research is also an opining mining; system using Analytic Hierarchy Process (AHP) of MCDM and DL to identify the public sentiment on different aspects of online teaching-learning from tweets during the pandemic. Several case studies have been conducted on the techniques proposed in this Dissertation based on different datasets to highlight the significance of this research.