Deep Learning Transformers for Sentiment Classification: A Performance Evaluation

Sentiment analysis, often referred to as opinion mining, is an important NLP technique used to detect and evaluate emotions conveyed in textual data. It plays a crucial role in understanding public sentiment across various domains, including social discourse, consumer feedback, and media analysis. This study examines the effectiveness of transformer-based deep learning models BERT (Bidirectional Encoder Representations from Transformers), RoBERTa, DistilBERT (a lightweight variant of BERT), and Electra in classifying sentiment within the Coronavirus Tweets NLP dataset.The objective is to assess and compare the performance of these models in accurately identifying the sentiment conveyed in tweets related to the Coronavirus pandemic. Model performance is primarily evaluated based on accuracy. Experimental results reveal that Electra outperforms BERT, RoBERTa, and DistilBERT, demonstrating its potential for more precise sentiment analysis on large-scale text datasets. These findings underscore the capabilities of transformer models in sentiment analysis and highlight their applicability to real-world NLP challenges.