CNN, LSTM, and Bi-LSTM based Self-Attention Model Classification for User Review Sentiment Analysis

Sentiment analysis, often known as opinion mining, is a computational job within the field of natural language processing (NLP) that entails the identification and classification of the sentiment or emotional disposition conveyed in a given text, such as user reviews. Several deep learning models are applicable for sentiment analysis, such as Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Bidirectional LSTM (Bi-LSTM) including self-attention processes. The use of sentiment analysis has become an essential tool for organisations and scholars that want to extract valuable insights from extensive amounts of textual data. The applications of this discipline are many and continuously developing, rendering it an area of active study and innovation within the wider domain and social environment of Natural Language Processing (NLP) and Artificial Intelligence (AI). This study focuses on the use of convolutional neural networks (CNN), long short-term memory (LSTM), and bidirectional LSTM with access to selfattention mechanism for sentiment analysis of user reviews. The CNN model exhibits an accuracy rate of 75%, while the LSTM model demonstrates an accuracy rate of 81%. Additionally, the Bidirectional-LSTM model showcases the highest accuracy rate among the three, reaching 84%.