Sentiment Classification for Big Data Applications Using an Ensemble Deep Learning Method

Abstract

This study proposes a computationally effective deep-learning model for the categorization of binary feelings to assess the polarity of sentiments portrayed in written text via people's ideas, attitudes, and emotions. One bidirectional long short-term memory (BiLSTM) layer and a global pooling method are used to accomplish this. Using a fusion gate R-CNN ensemble classifier, which can provide superior performance metrics than current models like LSTM, we expand our technique. The foundation of our idea is opinion mining, which is a pretty broad term. In our project, we have applied the idea of opinion mining to user comments. The polarity of the comments that are used to give expert commentary is determined by the use of CNN. Software interfaces describe the logical characteristics of each interface between the software product and the users, hardware interfaces describe the logical characteristics of hardware components, software interfaces describe the logical characteristics of each interface between the software product and other specific software components, communication interfaces describe the data items or messages entering and leaving the system, the services required, and the nature of communications, and hardware interfaces describe the logical and physical characteristics of hardware components. The project's distinctive feature is how it automatically creates expert comments after extracting user comments. Unlike an unregistered user, a registered user benefits from ideas that the system will create. Our design is computer-efficient and is suggested for real-time applications requiring sentiment analysis.