Natural Language Processing Techniques for Sentiment Analysis in Social Media

Abstract

This research has used the techniques of Natural Language Processing to explore how to do sentiment analysis on social media by using three algorithms, G-LSTM, RMDEASD, and BERT. It tested how the models perform on a different set of domains like disaster management, corporate performance, and consumer behavior for the purpose of sentiment classification. The study utilized more than 500,000 social media posts as the dataset, with accuracy rates of 88.4% for G-LSTM, 85.7% for RMDEASD, and 91.2% for BERT. Results reveal that BERT surpassed the other models in terms of accuracy and contextual understanding in aspect-based sentiment analysis. Moreover, the hybrid model G-LSTM was effective in disaster-related tweets by achieving a 92.3% accuracy in real-time sentiment classification. Comparisons with related work show that the proposed models significantly improve the accuracy and robustness of sentiment analysis over traditional machine learning methods. Challenges such as sentiment classification in low-resource languages are also addressed, providing insights on how to improve model applicability in diverse linguistic contexts. The findings add to the increasing body of research in sentiment analysis and indicate its potential applications in several industries, such as healthcare, marketing, and public opinion analysis.