Sentiment Classification of Tweets using Auxiliary Feature Integration with CNN and BiLSTM Networks

Nowadays, social networking sites are on the rise, generating a great quantity of data. Millions of individuals share their opinions on microblogging sites every day because they allow for quick and basic phrases. Opinion and sentiment analysis is an important job for identifying subjective information in social media postings. The purpose of sentiment analysis is to identify the views, feelings, and attitudes expressed in source information. Emotional sensations, such as dread, worry, or trauma, are typically the result of early psychological difficulties that may last a lifetime. Furthermore, individuals talk and share their thoughts on social media, frequently inadvertently expressing their secret feelings in the comments. In tweet sentiment analysis, views in tweets are often classified as positive or negative. This research presents a unique deep learning framework for tweet polarity recognition that integrates Auxiliary Feature Integration, Convolutional Neural Networks (CNN), and Bidirectional Long Short-Term Memory (BiLSTM) networks. The model uses auxiliary features to improve semantic comprehension, CNN to detect local text patterns, and BiLSTM to retain long-range relationships in both directions. Experimental evaluations on benchmark datasets show that the proposed model outperforms current techniques like CNN_GTO, GNN, KG-CNN, and Ensemble_Lexicon-based approaches in terms of accuracy, precision, and F1-score. The findings validate the model's capacity to handle a wide range of linguistic terms, making it ideal for real-world sentiment analysis applications using social media data. The comprehensive trials illustrate the framework's efficiency and indicate that transformer models outperform standard deep learning models. The framework strikes a compromise between accuracy and computing efficiency, making it appropriate for use in actual applications.