Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Sentiment Analysis on Online Product Review

Harshit Gupta1, Lokesh kr. Tiwari2, Kartik3, Lav Agarwal4, Atul Kumar5
1,2,3,4
(Bachelor of Technology, Final year student), 5Associate Professor,
Department of Computer Science & Engineering, IMS Engineering College, Ghaziabad, U.P., India

Abstract:- In today’s world opinion mining is widely II. METHODOLOGY

used topic, internet is having variety of data which is
rich in sentiments which are used by different The current research proposes knowledge system
companies for various purposes. Our aim is to develop a which aims at providing better review about the any
web application which embeds machine learning model product which is present on the e-commerce website. In
which provides the analysis of user reviews for this we take the comments given by the customers on the
particular product. product and review them and present them in such manner
that the user can understand the polarity of the comments
It shows the positive and negative polarity of on that product. These comments/reviews may be in the
reviews for searched product which will be helpful for form of phrases or sentences. To know the sentiment of the
users. In this application when a user searches for a users, these sentence have to be separated into Bag of
product, review data is collected from ecommerce words. These bag of words are checked with the trained
websites and that data is passed to a machine learning classifier bag of words and expressed these sentences into
model which is a naïve bayes classifier which classifies negative & positive words. In the e-commerce sites their is
the reviews into positive and negative sentiments based a high growth of data and these data is used to tell the
on the features extracted by the model. We show user quality of product. Now a days the popularity of these sites
the overall positive and negative polarity of reviews for are becoming high as they are major source for the
searched product and also we show how accurately we customers to predict the exact rating and reviews of the
have obtained the results. Thus these results helps user product.
to decide about the product.
So this the benefit of our system that it saves the time
Keywords:- Sentiment Analysis, Negative & Positive, and make it easy for the user to compile the rating and
Reviews of the Products. reviews and obtain the opinions from all this document,
providing the exact view of the product.
I. INTRODUCTION
III. IMPLEMENTATION
The purpose of this project is to develop website that
examines the product in positive and negative polarity on In this project we managed to organize the product
the basis of comments extracted from e-commerce website review into two categories - positive, negative. Below
given as input. Each sample will be processed for selective figure depicts the implementation process. The
features and an assessment will be done based on those implementation steps may include,
features in order to provide the right product.

 Sentimental Analysis
Sentiment analysis (sometimes referred as opinion
mining or emotion AI) it points towards the use of natural
language processing, analysis of text, computational
linguistic, & biometrics to systematically identify, extract,
quantify, and study important states and subjective
information.

The process of computationally distinguishing and

categorizing opinions indicates in a piece of text,
principally in order to determine whether the author's
approach towards a peculiar topic, product and many other
things is positive, negative, or neutral. It is context mining
of content which identifies and selects subjective
knowledge in source material, and helping a business to
interpret the social sentiment of their brand, product or
service while observing online conversations.

Fig 1:- Process of extracting features

IJISRT19MA710 www.ijisrt.com 712

Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
A. Data Collection D. Sentiment Classification
Data collection is the way to collect the basic data or
document on which work is done. Consumers express their  Naive Bayesian Classifier
sentiments about particular products on e-commerce Consider the case with one normally distributed
websites like amazons. Their sentiments and opinions are predictor and two classes. We classify to the highest input
expressed in different way, with different vocabulary, density, taking the prior into account. A generative
context of writing, usage of short forms and slang, making classifier is a model that specifies how to generate the data
the data huge and disorganized. Manual analysis of given the class conditional densities p(x|y = c) and the
sentiment data is virtually impossible. Therefore we uses (prior) class probabilities p(y = c). This is a model for the
the sentiment analysis to make this effort easy. joint distribution p(y, x). We compute the conditional
probabilities for classification using Bayes’ theorem,
The website is made with many products inside it and
the data about that products are collected in which reviews p(y = c|x) = p(x|y = c)p(y = c) P c 0∈Y p(x|y = c 0)p(y = c
and ratings of that products are stored. The assessment is 0) .
collected in the form of reviews which are given by the
different consumers of that product shows their opinion The Naive Bayes classifier (NBC) is a simple
about it. generative model based on the assumption that the
predictors are conditionally independent given the class
B. Pre-processing label. The class conditional density the becomes
The pre-processing is nothing but filtering the
extracted data before analysis. It includes identifying and p(x|y = c) = Y p j=1 p(xj |y = c).
eliminating non-textual content and content that is
irrelevant to the area of study from the data.

C. Feature Extraction
In the feature extraction we try to find out the best
suitable outcome from the reviews and ratings that are
generated about the particular product on the e-commerce
sites. To find out this we uses the method of bag of words
which can predict the result is best way.

Bag of Word is the technique in computer science

field known as natural language processing to extract
features from text. These features can be used for training
machine learning algorithms. It creates a vocabulary of all
the unique words occurring in all the documents in the
training set. The output of the bag of words model is a
frequency vector.

Fig 3:- This shows a comparison between the results on

model from sentiment calculated by taking rating we have
taken 3 star and above to be positive and 1 and 2 star are
taken as negative.

E. Data Set
In the data set it is distributed in 2 parts:

 Training data set : Training data set is taken from UCI

machine learning repository which contains 1000
reviews which are labelled positive and negative i.e. 1
and 0 . We have used 75% of data to train the classifier
and 25% to evaluate the performance of classifier.
 Second type of data is unlabelled data which are
reviews of various product which are to be classified as
positive and negative polarity by using our trained
classifier to review the product. We have collected these
Fig 2:- Different word count product reviews from Amazon.

IJISRT19MA710 www.ijisrt.com 713

Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
IV. RESULT B. Precision
Precision is the ratio of correct positive classifications
The result displays the report about the product based to the total number of predicted positive instances.
on the features(user review) and this report is generated on
the basis of general classification of positive or negative Precision = tp / (tp+fp)
polarity of the features given by the users on the website.
The percentage obtained during precision is 77.5%.
The link of the whole project where you can check
the whole working of the project with the code: C. Accuracy
Accuracy is the ratio of total number of correct
https://github.com/lokesh0412/product-review-using- predictions (true positive and negative) to the total number
sentiment-analysis.git of instances.

Accuracy = (tp+tn)/ (tp+tn+fp+fn)

The percentage obtained during accuracy is 76.4%.

D. F1 Score
F1 which is a function of Precision and Recall.F1
Score is needed when you want to seek a balance between
precision and recall.

F1= 2*[(precision*recall)/(precision+recall)]

The percentage obtained during f1 score is 75.91%.

Fig 4:-: Showing the polarity of different reviews

Fig 5:- Shows the percentage of positive and negative

reviews for product named cello ball pen
Fig 6:-: performance measure of a classifier
V. PERFORMANCE EVALUATION
VI. CONCLUSION
There are many ways in which you can obtain
performance metrics for evaluating a classifier and to Sentiment analysis is a area in which we finds about
understand how accurate a sentiment analysis model is. the consumer’s sentiments, their way of thinking or
Common methods used for evaluation include recall, emotions about the certain products. The problem on which
precision, accuracy and f1 score. this project works is the categorization of the sentiments in
the different polarity. In this project various reviews and
Abbreviations as follows: true positive (tp), false ratings of the different online product are taken from e-
Positive (fp), true negative (tn) and false negative (fn). commerce sites are stored and used as a basic data on
which classifier is imposed on reviews and ratings from
A. Recall which different bag of words are generated and using that
Recall is the ratio of correct positive classifications to bag of words polarity is decided. The sentiment generated
the total number of actual positive instances. from the website are shown in the best way so that
customer can easily understand the polarity of the reviews
Recall =tp/(tp+fn) generated on that e-commerce website. This is how we get
the best suitable product review.
The percentage obtained during recall is 74.4%.

IJISRT19MA710 www.ijisrt.com 714

Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

REFERENCES

[1]. Muhammad Taimoor Khan, Mehr Durrani 2,

Armughan Ali, Irum Inayat, Shehzad Khalid and
Kamran Habib Khan “Sentiment analysis and the
complex natural language” Khan et al. Complex Adapt
Syst Model (2016) 4:2
[2]. Kim S-M, Hovy E (2004) Determining the sentiment
of opinions In: Proceedings of the 20th international
conference on Computational Linguistics, page 1367.
[3]. Liu B (2010) Sentiment analysis and subjectivity In:
Handbook of Natural Language Processing, Second
Edition.. Taylor and Francis Group, Boca.Google
Scholar
[4]. Stanford (2014) Sentiment 140.
http://www.sentiment140.com/
[5]. M.Z.Asghar, A Review of Feature Extraction in
Sentiment Analysis, Journal of Basic and Applied
Scientific Research, 4(3)(2014), 181-186.
[6]. M.Eirinaki, S.Pisal and J.Singh, Feature-based opinion
mining and ranking, Journal of Computer and System
Sciences, 78(4)(2012), 1175-1184.
[7]. J.Isabella, Analysis and evaluation of Feature selectors
in opinion mining, Indian Journal of Computer Science
and Engineering, 3(6) (2012).

IJISRT19MA710 www.ijisrt.com 715