Synthesis of Slogans with Predicted Sentiment from Twitter using a Novel Hybrid SDG-LSTM Model for Election Campaigns

Objectives: The primary objectives of this study encompass the enhancement of election campaign strategies through the synthesis of sentiment-laden slogans derived from Twitter data. This is achieved by employing a novel Hybrid SDG-LSTM model, aiming to improve sentiment prediction accuracy and communication efficacy in the context of political campaigns. Methods: The process of slogan generation relies on sentiment prediction derived from sentiment-laden tweets. The proposed sentiment analysis methods for election campaign slogans encompass Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). A novel approach is introduced through the Hybrid SDG-LSTM model, leveraging the combination of Self-Distillation Guidance (SDG) with LSTM to enhance sentiment prediction accuracy and efficiency. This innovative method aims to provide a more robust and effective means of analyzing and generating slogans for election campaigns. Findings: The performance assessment of Deep Learning models, GRU, LSTM, and the Hybrid architecture, unveiled compelling outcomes. GRU showcased a commendable accuracy of 92.98%, while LSTM impressed with 95.91%. Remarkably, the Hybrid Spatial LSTM with GRU surpassed both, achieving perfection with 100% accuracy, precision, recall, and an exceptionally low loss of 0.0. These results underscore the superior performance and efficacy of the Hybrid model in sentiment analysis tasks. Novelty: The novelty of this research is encapsulated in the introduction of the Hybrid Spatial LSTM with GRU model, which demonstrates groundbreaking 100% accuracy, surpassing current models. This innovation capitalizes on the synergistic fusion of spatial attention mechanisms and the dynamic nature of GRU, marking a substantial advancement and establishing a new benchmark for highly accurate predictions in the domain of sentiment analysis.