Global Semantic Alite with Bidirectional Encoder Representations from Transformers for Sentiment Analysis

Sentiment analysis has gained significant popularity across various fields, particularly in the analysis of data collected from social media platforms like Twitter through Application programming interface (API).However, traditional sentiment analysis faces challenges, such as difficulty in capturing complex text structures and inefficiencies in handling large data, which negatively impact classification accuracy.This research proposes Global Semantic Alite with Bidirectional Encoder Representations from Transformers (GSAL-BERT) technique for efficiently analysing sentiments in Twitter data and enhancing classification accuracy.The model's global semantic component, combined with BERT, processes extensive text sequences by focusing on contextual relationships within sentences.The data used for this research is collected from the Twitter and Sentiment140 datasets.During pre-processing, stemming, Lemmatization, Stop word and removal are applied to reduce extraneous text and enhance the model's efficiency.Feature extraction using N-gram technique captures sequential word patterns, while Term Frequency-Inverse Document Frequency (TF-IDF) focuses on identifying the essential phrases for classification.The GSAL-BERT technique efficiently handles larger dataset and performs global sentiment analysis on Twitter data.The proposed method achieves a commendable accuracy of 0.9201 on Twitter data, and 0.9923 accuracy on the Sentiment140 datasets.When compared to the existing methods like Decision-based Recurrent Neural Network (D-RNN) and Long Short-Term Memory (LSTM) networks, the proposed model demonstrates an overall improved performance.