A Comparative Study of Sentiment Analysis Tools

Abstract

COVID-19 outbreak compelled people to stay at home due to complete lockdown in all the working areas. Immense use of World Wide Web and social media to exchange and share opinions, generated enormous web data to be utilized in the research work of the Natural Language Processing (NLP) field. Being a dominant side of NLP, Sentiment Analysis uses numerous tools to classify human sentiments as Positive (1), Negative (-1) and Neutral (0) so as to reach various conclusions. This research work focused on sentiment analysis of four datasets, web scraped from four different sources namely: Twitter, Facebook, Economic Times Headlines and news articles keyed by stock market. Seven contemporary and tremendously used sentiment analysis tools: Stanford, SVC, TextBlob, Henry, Loughran-McDonald, Logistic Regression and VADER are considered here to process four scraped datasets individually and analyses result in two ways: Facebook scraped data generates maximum overall positive sentiment score as 38.17% and VADER tool performs best among seven tools. VADER calculates overall positive sentiment score as 56.63%