Evaluating Sentiments in Social Media Comments on Tax Transformation in India using Deep Learning

The Goods and Services Tax (GST) was implemented by the government of India to have one tax for the entire country. Millions of taxpayers commented on their experience of the GST e-governance system on the Twitter platform, which was collated for the study from June 2017 to May 2020. This paper proposes a comprehensive approach for finding the variation of attention weights for key words (related to broad categories belonging to GST) present in the tweets over different quarters of the three-year period along with month-wise sentiment prediction for every quarter using Bi-directional LSTM model with attention. The contribution of key words, whose attention weights exceed two sigma thresholds, towards the net positive and negative sentiments of tweets is found to be significant in the study.