Author Email: firstname.lastname@example.org
Online shopping is a medium for many consumers to buy goods and services from E-commerce applications. The public is more inclined to online shopping from the last decade. These online shopping portals also provide the customers a way to express their experience with the product by giving reviews. These reviews not only prove a benefit to the consumer but also help the manufacturer in improving the quality of their products. Our proposed approach works on determining the polarity of the mobile phone’s reviews i.e. positive, negative or neutral using natural language processing techniques. We hope that our proposed model will act as a useful tool for online product review mining.
Nowadays online shopping has become a new trend of shopping. Myriad of people are attracted to this trending way of shopping. Online shopping portals avail the consumer of expressing their opinion about a product using reviews. These reviews not only help the general public in gaining the knowledge of the product but also the manufacturer to improve the quality of the product. Our proposed approach works on determining the polarity online product review using natural language processing tools. For example the review “The camera is awesome but the battery life is poor” here the reviewer has expressed positive opinion towards the camera but negative opinion towards battery-life. Our system works on determining fine grain polarity of the review. To achieve this, firstly we need to extract opinion target and opinion words from the review. The opinion target is the words about which an opinion is given. Like in this review words like “battery” and “camera” are opinion target. And opinion words are the words that express the opinion about the product. In the given example opinion words are “awesome” and “poor” our main aim is to determine the relation between the opinion targets and opinion words. Our proposed system has collected mobile review dataset from Amazon website. After that, we have identified opinion target and opinion words using rule-based pattern. Opinion words are assigned some weight. Some threshold value was decided, based on which the classification of the polarity was performed.
The proposed approach works on the classification of reviews of mobile phone taken from Amazon.com. We have used NLP techniques for extraction of opinion target and opinion words. We have tested the accuracy of different classifiers for the classification of the reviews. Figure 3 shows a graphical representation of the accuracy obtained from five different models. The accuracy obtained from Linear SVM is the highest i.e. 66.08. The second high accuracy is for Random Forest Classification i.e. 65.27. In the future, we would like to extend this work on some other product review and try to increase the accuracy of the proposed model
278 total views, 1 views today