Sentiment-Bearing String Extraction Method for Sentiment Analysis

Abstract

Consumers are increasingly being asked for comments, feedback or suggestions regarding the products they purchase by e-commerce platforms. This recent trend has led to a noticeable increase in the amount of online product reviews along with the significant growth in online shopping. Other potential consumers are strongly influenced by the recommendations or critiques shared by other fellow consumers while making purchase decisions. Sentiment analysis (SA) records every user's feelings, views, beliefs and opinion regarding the specified product in order to determine if the general mindset of the user is positive, negative or neutral. It enables potential customers to take purchase decisions and enterprises to make the necessary improvements to the product for better satisfaction of their consumers. In this paper, a method to extract sentiment bearing strings has been proposed for sentiment classification on Amazon Mobile review dataset which extracts the sentiment bearing strings from the review and classifies the review based on computed sentiment score. It applies four machine learning classifiers namely Random Forest (RM), Logistic Regression (LR), KNN and SVM. The experimental results show that by applying proposed method the classifiers achieve much improved classification performance in terms of precision, recall and F1-score and accuracy.The SVM achieves highest accuracy of 97.69%. The proposed method performed better in the experimental evaluation than the current approaches when it compared to others.