Analyzing Public Reactions to Budgetary Tax Reforms: A BERN-based Sentiment Study

Abstract

The Indian budget plays a significant role in the national economy, impacting different sectors and citizens. This study does a real-time analysis of public sentiment regarding the Indian budget through Twitter data. With the rise in the use of social media, it can be used for measuring public reactions. In this study, Python-based NLP methods are used to examine 9,780 tweets, classifying sentiment as positive, negative, or neutral. Robust preprocessing, such as removing stop words and special characters, was done to improve data quality. Sentiment polarity scores were computed by using a deep learning model based on BERT, which facilitates better context-aware sentiment classification. The BERT-integrated model showed a better accuracy of 92.84%, a 10.44% improvement over the 82.4% accuracy of the conventional lexicon-based methods. The findings reveal an even distribution of sentiments, with positive comments on social welfare policies and tax reforms, while fiscal deficits and inflation concerns yielded negative sentiments. This research highlights the efficiency of transformer-based NLP models in sentiment analysis, providing policymakers with useful insights for data-driven economic decision-making.