Sentiment Analysis on Social Media Using BERT Model

Abstract

The capacity to classify social media messages automatically regarding sentiment is increasingly crucial to understanding opinions of people. Current research advances prior sentiment detection investigations by revisiting performance of sentiment classification as per precision on social network datasets by utilizing BERT (Bidirectional Encoder Representations from Transformers) model. BERT model is utilized for improving sentiment classification performance on T4SA dataset in ongoing research and build on earlier unsupervised methods which have utilized manually constructed lexicons to detect extreme sentiments. Such findings display BERT poses potential to detect extreme emotions and improve classification performance significantly compared with past approaches while trained utilizing sufficient extreme sentiments dataset. Current investigation exposes probabilities of BERT for improving classification accuracy of social media sentiment detection, as part of advances in sentiment analysis. BERT‘s bidirectional encoding allows for accurate sentiment recognition by efficiently comprehending intricate verbal structures. The outcomes demonstrate BERT is a useful technique to examine public opinion, brand perception, and online discourse since it can distinguish subtle emotions in social media writing. Research concludes that BERT is an effective tool for sentiment analysis in social media because of its sophisticated contextual awareness and bidirectional encoding abilities.