DGDO-BiLSTM: Dominance Guiding Defense Optimization-based Bidirectional Long Short-Term Memory for Sentiment Analysis using Multilingual text and emojis

Abstract

Sentiment analysis plays an essential role in identifying someone’s emotional state, opinion, and perspectives, which in turn effectually utilized for obtaining product information and strategic decision-making process. However, the sentiment analysis exhibits some challenges, like performance degradation, difficult to categorize sentiment polarity, interpretation issues, and complexity problems. To resolve these drawbacks, proposed a Dominance Guiding Defense Optimization based Bidirectional Long short-term memory classifier (DGDO-BiLSTM) to evaluate the sentiment polarity of multilingual text and emoji classification. In this context, the DGDO-BiLSTM utilized Multilingual text and emoji-based review information to recognize the sentiments and attain certain information about the products. Further, the DGDO algorithm is utilized for enhancing the ability and efficacy of the model with the combination of Hippopotamus, and Walrus optimization algorithms, which effectually reduced the local optima issues and achieved an accurate convergence rate significantly. Meanwhile, the hybrid angular loss function is incorporated with the developed model to attain the superiority property and discriminative power that effectually minimizes the error rate gradually. Based on this effectiveness, the DGDO-BiLSTM model achieves better performance as 82.04 %, 95.31 %, 95.37 %, and 95.70 %, for negative predictive value (NPV), Accuracy. F1-Score, and Positive Predictive Value (PPV).