A Deep Learning based Model using Review Associated Feature Extraction Approach for Sentiment Analysis

Abstract

With the advancement of internet technologies, in the present days, online forums, social media platforms and e-commerce sites have made the product reviews process very easy. There are a lot of mobile applications, websites and forums where consumers used to share and circulate their opinions, experiences, ideas and views regarding products, brands and services. In consequence, online user reviews have become a deciding factor for many consumers prior to purchasing their selected items. The sentiment analysis is a technique to extract sentiments, feelings and insights from customer reviews and public texts. Therefore, plenty of businesses perform sentiment analysis in order to more thoroughly comprehend of their customer opinions and suggestions regarding their products and services. Furthermore, a number of scientific researchers also have a keen interest in classifying customer reviews into a set of labels employing text classification techniques. The objective of the this research work is to develop an approach to extract review associated features using Part-of-Speech (POS) tagging and design a CNN model to classify the reviews' sentiment as positive or negative. In this paper, an approach to extract review associated feature has been presented. Natural Language Processing (NLP) techniques are utilized for data preprocessing to remove uninformative data from reviews. Deep learning model CNN is used for sentiment classification and Amazon mobile reviews dataset is used for the experiment. The proposed model is experimentally evaluated and provides enhanced performance than other models also provides improved accuracy of 97.23% on Amazon mobile review dataset.