Sentiment Analysis of Remote Worker Tweets During COVID-19

In the midst of the COVID-19 pandemic, millions of people around the world faced the sudden shift to working remotely. This change triggered a range of positive, negative, and neutral reactions, which were visible on social media, notably Twitter. This study aims to analyze these complex sentiments to understand the public's perspective on remote working during the pandemic. We sourced extensive Twitter data from Kaggle and Data World platforms, which provided a comprehensive collection of tweets reflecting diverse public opinions. Utilizing this data, we applied three analytical tools, TextBlob, Vader, and RoBERTa, to examine the emotional content of each tweet. Vectorization techniques such as Bag of Words, TF-IDF, Word2Vec, GloVe, and BERT assisted in organizing the data, converting text into numerical form, and optimizing it for analysis with our machine learning models. The main purpose of machine learning classifications is to assess the performance of the sentiment analysis, thereby affirming the credibility of our results' credibility. Our analysis revealed that the amalgamation of RoBERTa, TFIDF, and the Stacking Classifier achieved a significant F1 score of 0.782. This high F1 score highlights the effectiveness of our model in accurately interpreting sentiments related to remote work during the COVID-19 pandemic. These findings underscore the criticality of adaptability and illuminate the essential contribution of real-time data in refining the remote work landscape molded by the ongoing impacts of COVID-19.