Sentiment Analysis on Omicron Tweets Using Hybrid Classifiers with Multiple Feature ExtractionTechniques and Transformer Based Models

Since the beginning of Covid-19, the world has been in a dilemma to cope up with its effects. With time the coronavirus has evolved into variants that caused a lot of destruction to human race. One such variant is “Omicron”. This variant made its presence in many countries throughout the world. The government is left in a straining situation to curb the spread of this variant and to stop the evolution of coronavirus. Though the strict precautions were exercised, the evolution was unstoppable. To understand the thoughts and feelings of the public, twitter can be considered as one of the best platforms for sentiment analysis. Analyzing the sentiments of people across the continents is horridly difficult but with the way technology has been making advancement in the world, analyzing has become a quiet easy job. In the existing studies on Covid-19, various word embedding techniques with machine learning and deeplearning classifiers has been used for the analysis. Language based models have proven to achieve higher accuracy forsentiment analysis. Amidst these hybrid classifiers, have performed tremendously good. In the proposed work, seven Machine Learning hybrid classifiers are compared with four single classifiers using TF-IDF and Word2Vec. A proposedDeep Learning hybrid classifier is compared with two single classifiers using GloVe and FastText. Furthermore, language models like BERT and RoBERTa are employed in an effort to boost validation outcomes upto 93.39% and 93.47%.