Categorizing the tweets are difficult based on their simplicity and use of standard or non-standard slang. There have been several types of research that have discovered highly precise sentiment data classifications, but very few of them have been validated on Twitter data. Techniques of sentiment analysis in the past have mostly dealt with interpreting binary or ternary emotions in unilingual texts. Furthermore, emotions become apparent in writings that are written in more than one language. Today's society increasingly relies on the usage of emotions and summarise key points in written communication, such as tweets on the Covid Vaccine. Earlier machine learning methods focused exclusively on text classification, image classification, or emotion classification that ignores the majority of emotions. Based on the examination of text and emotions in tweets on the Covid Vaccine Sentiment Analysis (SA), this paper propose improved SVM- Fuzzy using BERT technique together to enhance the performance measures. The proposed Improved SVM- Fuzzy method is to improve the performance measures (accuracy) and reduce the error rate based on feature extraction and classification in sentiment analysis. In experimental circumstances, the proposed model outperformed state-of-the-art approaches in tweet emotion identification and text recognition by 97.3%.

BERT, SVM- Fuzzy, Twitter Data, Sentiment Analysis, Emoji.