Mining Online Opinions and Reviews Using Bi-LSTM for Reputation Generation

Nowadays, people express their opinions and thoughts about an item (products, movies, services, and brands, etc.) directly online. They use platforms like IMDB, Amazon, eBay, and more. This large amount of people’s opinions is very valuable for an automatic computing of reputation score for companies. Consequently, developing a system able to generate reputation from textual opinions and their attached rating will be a helpful tool for companies to have an idea about the quality and the issues of their product, and for potential customers to assist them during their ecommerce decision-making. However, a very limited number of studies have investigated mining reviews expressed in natural language for reputation generation. Therefore, we propose an approach and a whole system for generating reputation using bidirectional long short-term memory Recurrent Neural Network (Bi-LSTM RNN) and Natural Language Processing (NLP) techniques. The experimental results conducted on the IMDB dataset show that our method provides the nearest reputation value to the ground truth (IMDb weighted average vote). This implies that the proposed approach can be applied in practice to generate reputation.