fProSentiment Analysis on Mobile Phone Brands Reviews using Convolutional Neural Network (CNN)

Abstract

Due to the rapid growth of online e-commerce, customers can now voice and express their reviews and thoughts on online products. Therefore, companies who market their product on the e-commerce website will receive thousands of reviews and feedback from their end-users directly on this platform. As the amount of textual data grows tremendously, developing sentiment analysis that automatically analyses text data becomes increasingly vital. It is because reading every review manually can be a tedious task and time-consuming. Analyzing the sentiment for all reviews can provide the companies with an overview of how positive or negative the customers are about their products. The convolutional neural network (CNN) has recently been used for text classification tasks and has achieved impressive results. Hence, this study proposes a CNN method for sentiment analysis to classify the reviews on mobile phone brands. The customer reviews dataset is obtained from the Amazon website. This study combined the Word2Vec-CNN model to predict the sentiment of mobile phone reviews effectively. Pre-trained Word2Vec model is utilized to generate word vectors in word embedding. CNN layers are used to extract better features for sentence categorization to identify the sentiment polarity of the reviews, whether positive or negative. The obtained results give us 88% accuracy and the developed application can also function well in analyzing the sentiment of customers’ reviews.