SENTIMENT ANALYSIS OF PRODUCT REVIEWS

Abstract

During the past decades, online market places have been popular and most of the sellers also request the customers to express the reviews of the products. Nowadays individual and groups depend heavily on website for consumers’ reviews in their agreement on buying the product. User generated opinioned data are increasing day by day as consumer left opinions about the product they bought. Product manufacturers also need to take time for analyzing the huge amount of opinions. With the increasing amount of text data, sentiment analysis is becoming more and more important. Sentiment analysis is commonly used with Natural Language Processing. This paper expresses about the sentiment analysis, which is the process of mining the texts, in order to distinguish the extract written by the user. So, the paper proposes a framework for reviews data using hybrid approach used in lexicon and machine learning approach to classify the review text whether they are positive opinion, negative opinion, and neural opinion. The approach describes a guideline for training data using Vader lexicon and testing data using machine learning algorithm and demonstrates the classification approach of supervised learning using Multinomial Naïve Bayes on Amazon product review dataset. The paper presents the evaluation results as positive reviews are found the most and negative reviews are found the least.