Polarity and Subjectivity Analysis for Social Media Sentiments Amidst COVID-19

Textual content in the form of messages, texts, tweets, comments, etc. is generated on a daily basis on various social media platforms. The foremost objective of this paper is to work on social media comments to evaluate polarity and subjectivity of the created corpus during COVID-19. A novel CPA: Context Based Polarity Analysis and an ADDT: Annotated Dependency based Decision Tree are used to achieve classification and sentiment score for the corpus. The findings of the work include disclosure integration and pragmatic results that in turn provide decision for sentiment behind the corpus. The results can be used to gain insights to researchers working on Sentiment Analysis and authorities concerning to current pandemic. The proposed work is also compared with some state of art findings by other researchers and is believed to have produced better results in the field of Polarity and Subjectivity of a corpus based text analysis on the basis of three parameters i.e. Document Level accuracy, Sentence Level accuracy and Aspect Level accuracy. To achieve this, NLP is practiced to extract sentiments and comparatively evaluate them on three major online social media platforms.