Opinion Mining and Sentiment Analysis of Reviews

Introduction

In the recent past, e-commerce sites have made a rapid growth and the number of customer reviews that a product receives have grown rapidly.This makes it difficult for a potential customer to read and analyze these reviews because of their large number and unstructured nature.One of the possible solutions to the above problem could be a tool that can automatically summarize the product review based on user opinion.

Thier is a great need dynamic system that can mine the features of the product on which the customers have expressed their opinions and perform summarization accordingly.These type of systems are really helpful for the consumer as well as for the manufacturer of the product, as it becomes easy for the manufacturer to keep track and to manage customer opinions.These systems should be highly accurate and should cover every aspect of the reviews so that customer gets an unbaised form of summarization.

Approach

The system could be broadly categorized into three parts:-

Feature-Opinion mining

It is the process of identifying/finding user opinion about a product.This part of the system is only concerned with mining the product feature and user opinion from the unstructured reviews.In this phase, the reviews get broken in the feature-opinions list.

Sentiment analyses

Analyzing the sentiments of the user from the extracted feature-opinion list is termed as sentiment analysis.In sentiment analyses, the user opinions are categorized into negative, positive or neutral.

Summarization

This part of the system is only concerned with the summarization of reviews based on user opinion.

Techniques

The system is dived into three parts and there are the different set of techniques available for each phase of the system.

Feature-Opinion minning

POS(parts of Speech Tagging) is one of the famous technique used by many researchers for mining the feature-opinion list from the product reviews.

The POS help in classifying the position of each word in the sentence.

“Display is superior and of top-notch quality”

 POS tagging of the review:-

(‘Display’, ‘NNP’), (‘is’, ‘VBZ’), (‘superior’, ‘JJ’), (‘and’, ‘CC’), (‘of’, ‘IN’), (‘top-notch’, ’JJ’), (‘quality’, ‘NN’).

some other techniques like naive Bayes classification is also used for opinion mining and text classification

sentiment analysis

Many research used a list of positive words and a list of negative words for analyzing the sentiment of user review.Various other techniques where objective sentences were removed by extracting subjective ones are also used for sentiment analysis of reviews.

Summarization

Extraction based summarization, Abstraction based summarization and Aided summarization are different techniques for the summarization however extraction based technique is widely used.

Recently,Many research have also tried using some deep learning based technquies for Opinion Mining and sentiment analysis of customer reviews. The deep learning model have outperformed all the traditional model where the opinion mining and sentiment analysis are perfomed using a constrained model.