Enhancing Sentiment Analysis Performance Using Deep Learning Techniques

Sentiment analysis plays a big and important role in numerous domains like feedback of customer, monitoring of social media, research in market, and many more. However, the sentiment analysis solutions such as lexicon-based method, SVM, Naïve-Bayes, RNN, CNN based method etc., even though they work to some extent, are easily puzzled by nuances of language and never completely maintain contextual data. This paper discusses the hybrid sentiment analysis methods using deep learning techniques which improve the accuracy. This paper offers a comprehensive literature review examining traditional approaches and deep learning approaches applied to sentiment analysis. This paper then explores the hybrid approaches for sentiment analysis namely bidirectional LSTM with attention mechanism and transformer-based approach, which are specifically developed to enhance the accuracy and perform the experiments to test these two proposed approach's efficacies with several benchmarks. As a result, these hybrid approach clearly exceeds the previous traditional methods, demonstrating the utility of using deep learning for sentiment analysis tasks.