A modernized approach to sentiment analysis of product reviews using BiGRU and RNN based LSTM deep learning models

Area of Science:

• Natural Language Processing
• Deep Learning
• E-commerce Analytics

Background:

• The rise of Web 2.0 and online shopping has led to a surge in user-generated content.
• Companies increasingly use opinion mining to analyze market demands and product reception.
• Sentiment analysis is crucial for understanding consumer opinions in the e-commerce industry.

Purpose of the Study:

• To propose a novel sentence-level sentiment analysis approach for mining online product reviews.
• To develop a model that preprocesses, extracts features, and classifies sentiment polarity.
• To evaluate the model's performance against existing techniques.

Main Methods:

• Web crawling for product review collection.
• Natural language processing techniques for preprocessing (tokenization, lemmatization, stop word removal, named entity recognition, part of speech tagging).
• Bidirectional Gated Recurrent Unit (BiGRU) for feature extraction.
• Hybrid Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) for sentiment classification (positive, negative, neutral).

Main Results:

• The proposed model achieved promising performance metrics (accuracy, precision, recall, F-measure, AUC).
• Comparison with existing models like Deep Convolutional Neural Network (CNN), Multilayer Perceptron (MLP), CapsuleNet, and Generative Adversarial Networks (GANs) showed superior results.
• The novel combination of techniques and application to online product reviews demonstrated effectiveness.

Conclusions:

• The developed sentiment analysis model offers a novel and effective approach for analyzing online product reviews.
• The model has broad applicability in market research, social network mining, recommendation systems, brand analysis, and product quality management.
• The findings suggest significant potential for deep learning and NLP in enhancing e-commerce intelligence.