Event Based Sentiment Trend Analysis using Deep Learning Techniques

Abstract

The ever-growing volume of social media data presents a unique opportunity to understand public opinion on various topics. This study focuses on sentiment trend analysis, which uses social media text to uncover evolving emotional patterns. By analyzing these trends, researchers can gain valuable insights applicable in diverse fields. The research proposes a framework for analyzing user tweets related to specific events, outlining steps like data retrieval, event prompts, data segregation, preprocessing, and sentiment analysis using deep learning (BiLSTMs and BERT), model evaluation, and sentiment trend analysis. The experiment uses Twitter data for events like the COVID-19 pandemic, Russia-Ukraine war, and IPL (Indian Premier League). Results show the effectiveness of the approach, with sentiment analysis models achieving over 90% accuracy in some cases. Sentiment Trends reveal predominantly negative sentiment surrounding the Russia-Ukraine war and positive sentiment associated with the IPL.