A Benchmark Study in Sentiment Analysis with Deep Neural Networks

Sentiment analysis plays an increasingly crucial role in gaining insight into people’ perceptions of particular firms, products, or entities-whether they be virtual or physical-as text data continues to rise exponentially. Sentiment analysis enables us to methodically assess user evaluations and makes conclusions based on the thoughts stated in these reviews as a whole. Within the field of deep learning, popular architectural models used for sentiment analysis projects include the recurrent neural network (RNN), convolutional neural network (CNN), long short-term memory (LSTM), bidirectional LSTM (Bi-LSTM), and gated recurrent unit (GRU). For the purpose of this investigation, a thorough comparison of the LSTM, Bi-LSTM, RNN, CNN, and GRU architectures was carried out for sentiment classification, using the IMDB movie reviews dataset to determine which architectural setup was best. Our results clarified that GRU had the quickest training time and outperformed other models when it came to accurately classifying emotions in IMDB movie reviews.