The increasing ubiquity of Internet usage has made sentiment analysis an essential field of research in the rapidly developing field of natural language processing. This research makes a new contribution to sentiment analysis during the COVID-19 pandemic by looking at the emotional responses expressed in comments and tweets. Emotions are extracted from a large set of annotated tweets (March 23, 2020, to July 15, 2020) using content mining. A variety of methods, such as sentiment lexicon-based approaches, topic modeling, and keyword extraction, are used in content mining. Several content mining approaches are used to extract emotional feelings from the tweets after they have been pre-processed to remove noise and extraneous information. Bidirectional and Auto-Regressive Transformers (BART) and Long Short-Term Memory (LSTM) models increase the accuracy of emotion classification. Comparing models reveals COVID-19 Twitter emotions for informed outbreak control. Consistency is ensured by human annotators who classify tweets into emotion categories such as happiness, grief, fury, fear, disgust, and surprise. Annotation efficiency can be improved with automated sentiment tools.