Sentiment Analysis Based on Deep Learning in E-Commerce

Abstract

Social media allow businesses to find out what customers are thinking about their products and to participate in the conversation. Companies, therefore, have an interest in using them to market their products, identify new opportunities and improve their reputation. The main objective of our study was to recognize feelings expressed in opinions, ratings, recommendations about a product using a construction based on a corpus of sentiment lexicon with different deep learning algorithms. In this work, we will then analyze an e-commerce platform in order to know the feelings of customers towards the products. This study is conducted based on a static dataset of 41,778 smartphone product reviews in french collected on Amazon.com. For the classification of reviews, we applied the Long short-term memory network (LSTM). The results showed that the LSTM deep learning algorithm yielded a good performance with an accuracy of 95%.