help

bcs

Automatic_sentiment_analysis_of_user_reviews.pdf

Home >Computer Science homework help >help

2016 IEEE International Conference on Technological Innovations in ICT For Agriculture and Rural Development

(TIAR 2016)

158

Automatic Sentiment Analysis of User Reviews

Abinaya.R1, Aishwaryaa.P2, Baavana.S3, UG Students,

Thamarai Selvi.N.D.4, Assistant Professor, Department of Computer Science and Engineering,

Easwari Engineering College, Chennai, India. [email protected], [email protected] ,

[email protected], [email protected]

Abstract — Data mining is the process of turning raw data into

useful information. The main use of data mining is to fetch the

required data and extract useful information from the data and

to interpret the data. In the existing system, Bag of Words model

is used along with Dual sentiment Analysis in order to classify

the reviews as positive, negative and neutral. However, the

performance of Bag of Words sometimes remains limited due to

some fundamental deficiencies in handling the polarity shift

problem. The proposed system uses a dictionary based

classification for accurately classifying the reviews as positive,

negative and neutral. To enhance the accuracy in the

classification of neutral reviews, Support Vector Machine

algorithm is implemented. Both the product owner and the user

can identify the quality of the product based on the sentiment

graph that is generated based on the reviews for each of the

product video. A comparative study of the sentiment graphs is

performed in order to improve the efficiency of visual

representation.

Keywords: Opinion Mining, Sentiment Analysis, Data Mining

I. INTRODUCTION

Internet has gained its access in each and every

domain and has become an integral part in all walks of life. In today’s world, people give their reviews over the internet.

All these reviews are being stored as huge amount of data for

each and every minute. These reviews have an equal

importance as that of buying the commodity. Analyzing these

reviews will enhance both the end users and the account

executive. Sentiment analysis has earned its identification and

is used in classifying the reviews.

Sentiment analysis, also called opinion mining, is the

field of study that analyses people’s opinions, sentiments,

evaluations, appraisals, attitudes, and emotions towards

entities such as products, services, organizations, individuals,

issues, events, topics, and their attributes. It represents a large

problem space. There are also many names and slightly

different tasks, e.g., sentiment analysis, opinion mining,

opinion extraction, sentiment mining, subjectivity analysis,

affect analysis, emotion analysis, review mining, etc.

However, they are now all under the umbrella of sentiment

analysis or opinion mining.

While in industry, the term sentiment analysis is more

commonly used, but in academia both sentiment analysis and

opinion mining are frequently employed. They basically

represent the same field of study. The meaning of opinion

itself is still very broad. Sentiment analysis and opinion

mining mainly focuses on opinions which express or imply

positive or negative sentiments.

To do an analysis, classification plays a key role in

opinion mining. A Classification Algorithm is a procedure for selecting a hypothesis from a set of alternatives that best fits a

set of observations.

Opinions are central to almost all human activities

because they are key influencers of our behaviours. Whenever

there is a need to make a decision, others’ opinions are

required. In the real world, businesses and organizations

always want to find consumer or public opinions about their

products and services. Individual consumers also want to

know the opinions of existing users of a product before

purchasing it, and others’ opinions about political candidates

before making a voting decision in a political election. In the

past, when an individual needed opinions, he/she asked

friends and family. When an organization or a business

needed public or consumer opinions, it conducted surveys,

opinion polls, and focus groups. Acquiring public and

consumer opinions has long been a huge business itself for

marketing, public relations, and political campaign companies.

Opinion summarization summarizes opinions of articles

by telling sentiment polarities, degree and the correlated

events. With opinion summarization, a customer can easily

see how the existing customers feel about a product, and the

product manufacturer can get the reason why different stands

people like it or what they complain about.

A seller’s job can be quite complicated or it can be quite

easy. The two contradictory terms define the selling

experience, based on the fact as how seller interprets the

consumer interests. Unless one is a psychic or knows how to

get into others mind the actual demand of the customer’s and

the product can’t be collaborated. Having a right product is

important and equally important is to present it before the

2016 IEEE International Conference on Technological Innovations in ICT For Agriculture and Rural Development

(TIAR 2016)

159

right customer (one who actually needs it or is interested in it).

The product should put on positive feeling of ownership among the individuals. And such feelings are clearly

expressed in opinion mining polls.

II. LITERATURE SURVEY

Rui Xia, FengXu, Chengqing Zong, Qianmu Li, Yong Qi,

and Tao Li [10].In recent years, with the growing volume of

online reviews available on the Internet, sentiment analysis

and opinion mining, as a special text mining task for

determining the subjective attitude (i.e., sentiment) expressed

by the text, is becoming a hotspot in the field of data mining

and natural language processing [2], [7], [8]. Sentiment

classification is a basic task in sentiment analysis, with its aim

to classify the sentiment (e.g., positive or negative) of a given

text. The general practice in sentiment classification follows

the techniques in traditional topic-based text classification,

where the Bag-of words (BOW) model is typically used for

text representation. In the BOW model, a review text is

represented by a vector of independent words.

Although the BOW model is very simple and quite

efficient in topic-based text classification, it is actually not

very suitable for sentiment classification because it disrupts

the word order, breaks the syntactic structures, and discards

some semantic information. Consequently, a large number of

researches in sentiment analysis aimed to enhance BOW by

incorporating linguistic knowledge. However, due to the

fundamental deficiencies in BOW, most of these efforts

showed very slight effects in improving the classification

accuracy. One of the most well-known difficulties is the

polarity shift problem. Polarity shift is a kind of linguistic

phenomenon which can reverse the sentiment polarity of the

text. Negation is the most important type of polarity shift.

A simple yet efficient model, called dual sentiment

analysis (DSA), is developed to address the polarity shift

problem in sentiment classification. By using the property that

sentiment classification has two opposite class labels (i.e.,

positive and negative), we first propose a data expansion

technique by creating sentiment reversed reviews. The

original and reversed reviews are constructed in a one-to-one

correspondence. Thereafter, we propose a dual training (DT)

algorithm and a dual prediction (DP) algorithm respectively,

to make use of the original and reversed samples in pairs for

training a statistical classifier and make predictions.

To reduce DSA’s dependency on an external antonym

dictionary, we finally develop a corpus-based method for

constructing a pseudo-antonym dictionary. The pseudo

antonym dictionary is language-independent and domain-

adaptive. It makes the DSA model possible to be applied into

a wide range of applications.

III. PROPOSED SYSTEM

The system consist of three major steps

1. Reviews Extraction i. Creation of user interface and uploading videos

ii. Word clustering iii. Preprocessing

2. Reviews Evaluation i. Selection of hypotheses set

ii. Classification

3. Sentiment graph visualization.

The first step of the module involves creation of web based

interface. Upon creation, the admin will login and upload

videos which are in MP4 format. Then, the user will login and

view videos. He shall then give reviews for the video which

will get stored in the database. The stored reviews are then

processed with the help of word clustering and pre processing.

In review evaluation, refining of the text in the form of

relevant phrases, words with the help of classification is

performed. Classification is based on the supervised learning.

The algorithm classifies the words given by the user into

positive, negative and neutral. The words that do not have any

meaning is considered as a hypothetical words. Such kind of

words is processed with the help of an iterative induced

hypothesis. The last module involves generation of a

sentiment graph. It takes the classified reviews that are stored

in the database as a input which are visually shown as the

sentiment graph. The Figure 1 describes the functional

architecture of the proposed system.

2016 IEEE International Conference on Technological Innovations in ICT For Agriculture and Rural Development

(TIAR 2016)

160

Fig.1 Proposed System

IV.IMPLEMENTATION

The implementation of the system involves the following

modules:

A. Review Extraction

• Creation of Web based Interface.

• Extracting words from text using Word Clustering

that influence the outcome of the result.

• A preprocessing of words are executed to avoid

duplication in words.

Pseudocode: Word Clustering

Given: a dictionary D of size N

Create: table of clusters

i=0

while (D is not empty)

G1=D[0]

t=sizeof(D)

for j from 1 to t

if d1(G1,D[j])=1 then d1(G1,m1)=1<=i<=n D[j]->M1

end for

remove G1 and all M1 members from D

(G1,M1)->C

i=i+1

end while

Return C

Pseudocode: Preprocessing

Sql query:

select tb1.a1,tb2.a2 from tb1,tb2

from tb2 inner join tb1 on tb1.id=tb2.id

where tb1.a1=“”

B. Review Evaluation

• It involves refining text in form of relevant phrases,

words, etc., which are contained in a set called

hypothesis space.

• Induced Hypothesis is used for those words that are

to be classified by the owner of the video.

• Supervised learning classifies the words into positive,

negative and neutral which is implemented in coding

part.

• Architecture for Support Vector Machine Learning is

shown in Figure 2.

Fig.2 Support Vector Machine Learning

 A sample of reviews and their class to which it is

classified into, is shown in the Table 1 for all types of

sentiments.

Table1

Reviews and their class

Review Text Class

Very Nice Positive

Great!!! Neutral

Very Bad Negative

Average Neutral

WOW!* Neutral

C. Sentiment Graph Visualization

• It involves producing a sentiment graph for the

classified reviews which are useful in a way of

understanding the product.

Pseudocode:

video(){

con.Open();

2016 IEEE International Conference on Technological Innovations in ICT For Agriculture and Rural Development

(TIAR 2016)

161

SqlCommand cmd = new SqlCommand("select

count(id) as vid,count(Status) as Status from reviews

where vid='" +

Convert.ToInt32(Session["id"].ToString()) + "' and

Status='Positive'", con);

SqlDataAdapter dr = new SqlDataAdapter(cmd);

DataTable dt = new DataTable();

dr.Fill(dt);

Chart.DataSource = dt;

Chart.DataBind();

con.Close();

V. RESULTS AND PERFORMANCE ANALYSIS

A. Classification of sentiment words

In the classification of the sentiment words, the reviews

given by the user is classified as positive, negative and neutral

by using Support Vector Machine Algorithm as shown in

Figure 3.

Fig.3 Classified reviews

The process takes place by the reviews being parsed and

traversed between database and the dictionary which will

define the sentiment extraction patterns for the predicates

mentioned.

The reviews given by the user may be sentence level,

word or a document level and proper classification will be

done and the score of the reviews are inferred with colors,

where positive is represented by green, negative is represented

by red and neutral is represented by yellow. The neutral words

may have some hypothesis words present in it. These

hypotheses are again classified using the Supervised Learning

algorithm. This algorithm will classify the words recursively

until proper score is obtained.

B. Sentiment Graph Visualisation for Sentiwords

A new approach called sentiment graph is introduced.

This shows the classification of the reviews in the form of

graph as shown in Figure 4

Fig.4 Sentiment Graph

The process of computationally identifying and

categorizing opinions expressed in a piece of text, especially

in order to determine whether the writer's attitude towards a

particular topic, product, etc. is positive, negative, or neutral.

Here the classification of reviews is represented using the

bar graph. There exist a practical inefficiency wherein

executive or users need to go through all the comments to find

out the number of positive, negative, neutral comments before

coming to a decision. The Sentiment graph reduces this

difficulty.

The graph has X-axis and Y-axis, wherein the X-axis

represents the sentiment field and Y-axis represents the

number of reviews given by the user. This information will

create a major revolution in marketing and advertising field.

C. Comparison of the Sentiment graph

The different videos uploaded by the admin are viewed by the users and then reviews given by them are

compared for better understanding as shown in Figure 5

2016 IEEE International Conference on Technological Innovations in ICT For Agriculture and Rural Development

(TIAR 2016)

162

Fig.5 Comparison of Sentiment Graph

With the help of the comparison mentioned in the graph

the release and the sales of the product can be determined. The

accuracy of the classification is also improved in comparison.

The clear picture of the sentiment classification is achieved

here.

In this proposed system comparison between the two

systems can be viewed by the user at the same time and proper

perception and feedback for the videos uploaded by the admin

is obtained. This the major result provided in this system

which is not present in the existing system.

VI. CONCLUSION

Sentiment analysis is essential for anyone who is going to

make a decision. Sentiment analysis is helpful in different

field for calculating, identifying and expressing sentiment. It

is helpful for everyone when they want to buy a product and

they can decide which product is best. Sentiment analysis is

very important for Enterprises and helps them to know what

customers think about their products. Therefore companies

can take decisions about their products based on customer’s

feedback Thus companies can modify their products features

and introduce new products according to customer’s opinion

in a better and a faster way.

VII. FUTURE ENHANCEMENT

There is a lot of scope in analyzing the video and the

images on the web. Nowadays with the advent of Facebook,

Instagram and video vines, people are expressing their

thoughts with pictures and videos along with text. Sentiment

analysis will have to pace up with this change. Tools which

are helping companies to change strategies based on Facebook

and Twitter will also have to accommodate the number of

likes and re-tweets that the thought is generating on the social

media. People follow and unfollow people and comments on

social media but never comment. So there is a scope in

analyzing these aspects of the web as well.

REFERENCES

[1] Abbasi,A.H. Chen, and A. Salem, “Sentiment Analysis in Multiple

Languages: Feature Selection for Opinion Classification in Web

Forums,” ACM Trans. Information Systems, vol. 26, no. 3, pp. 1-34,

2008.

[2] Abbasi, S. France, Z. Zhang, and H. Chen, “Selecting attributes for

sentiment classification using feature relation networks,” IEEE

Transactionson Knowledge and Data Engineering (TKDE), vol. 23,

no. 3, pp.447-462, 2011.

[3] Chenghua Lin, Yulan He, Richard Everson, Member, IEEE, and

Stefan Ru¨ger, June 2012,”Weakly Supervised Joint Sentiment-Topic

Detection from Text”, IEEE transactions on knowledge and data

engineering, VOL. 24, NO. 6

[4] Dasgupta,S. and V. Ng, “Topic-Wise, Sentiment-Wise, or Otherwise?

Identifying the Hidden Dimension for Unsupervised Text

Classification,” Proc. Conf. Empirical Methods in Natural Language

Processing (EMNLP), pp. 580-589, 2009.

[5] Liu,B. “Sentiment Analysis and Opinion Mining,” Synthesis Lectures

on Human Language Technologies, vol. 5, no. 1, pp. 1-167, May

2012.

[6] Pang,B. and LeeL,.“Opinion mining and sentiment analysis,”

Foundations and Trends in Information Retrieval, vol. 2, no. 1-2, pp.

1-135,2008.

[7] Rui Xia, FengXu, ChengqingZong, QianmuLi, Yong Qi, and Tao Li,

August 2015,” Dual Sentiment Analysis: Considering Two Sides of

One Review”, IEEE transactions on knowledge and data engineering,

Vol.27, AnNo.8.