International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)

ISSN: 0976-1353 Volume 22 Issue 2 – MAY 2016.

POLARITY IDENTIFICATION THROUGH EMOTICON

USING CONTEXT BASED SENTIMENT ANALYSIS
Priya C#1, Sankar Ganesh K#2
#1
Department of computer science and engineering, Kingston engineering college, Vellore
[email protected]
#2
Department of computer science and engineering, Kingston engineering college, Vellore
[email protected]
Abstract:-The use of emoticons in text to express opinion and to use )I attempt to pinpoint the words with the polarity and
sentiment has become more and more common. So emoticons classify them as happy, unhappy, neutral and mine the opinion
play an important role in conveying a text's overall sentiment. in large amount of data in less time in good consistency.
This project aims to extract the meaning of the emoticons used in
C. Proposed System:
context of the message and provide a more clear and accurate
The proposed system will analyze the emoticons used
analysis of the sentiment of the text.
in the message and determine the sentiment of the text. first I
Index Terms: Emoticon, Context, sentiment analysis, opinion
identify the meaning of the emoticon .Then to understand it in
mining. the context of the text and I analysis where it is placed, how it
I. INTRODUCTION: affects the text, what it expresses and relate it with the
The sentiment analysis plays a Vitol role nowadays message to provide a clear view of what the person is trying to
every human being are interested in knowing “what other’s convey. From the live dataset from the twitter is extracted and
think?” “What might be their opinion about it?” And very
then the opinion is given since the r-language is the data
recently we all look into the ratings and read the reviews or
feedback given in it and decide to choose the correct one. analysis process the r-language is used to read multiple dataset
Sentiment analysis is the process of computationally and make the process of machine learning easy even when
identifying and categorizing opinions expressed in a piece of there is ton’s of data. And not only it makes the machine
text, especially in order to determine whether the writer's learning process easy but also gives the accuracy in providing
attitude towards a particular topic, product, etc. Is positive, the results and increases the efficiency when compared with
negative, or neutral. This sentiment analysis and the opinion
the existing system and also recognizes the emotions and gives
mining are like similar they reduce the time of reading the
reviews and feedback and as well as compares the rating and the result in which level of polarity is that depended to.
the review and provide the efficient opinion whether it is II. ARCHITECTURE
positive or negative or neutral comment. This not only The architecture diagram shows the step by step
categorizes the polarities but also classifies the emotion such process that how the twitter is authenticated and then how it
as sad, happy, disgust, fear, joy and shows the level of authorizes the developer and then how the search term entered
polarity’s emotion. So far the sentiment analysis and opinion in to the twitter. After all those authorization given to the user
mining used was not showing the accurate result and does not
the tweets of current 8 days are being extracted and then the
recognize the emoticon in it. Here I have improved the
accuracy, efficiency, and made emoticons to be recognized. preprocessing and the subjectivity analysis done and then the
A. Existing system: opinion are given based on the search term.
Current sentiment analysis approaches typically do
not consider emoticons. On the contrary, one of the first steps
in most existing work is to remove many of the typographical
symbols typically constituting emoticons, thus preventing
emoticons from being detected at all. Even in systems that
identify emoticons, they are considered as standalone entities
and not analyzed in context to the message.
B. Problem statement:
As analyzed the existing system, the opinion mining
from content based emoticon are not done up to the accuracy
and be consistent. In this paper I focus on the identification of
polarity in the text and the emoticon(e.g., this mobile is terrific

396
 International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353 Volume 22 Issue 2 – MAY 2016.

• Removes retweets
• Removes emoticons
• Removes punctuations and URL
That is it will make the unstructured data in to the
structured format so the machine will read only the original
text and provide the result and also it reduces the time of
extracting the context from the emoticons. because it first
finds the opinion of the context in the feedback or review and
the emotion that which is inside the context is analyzed easily
and later only it will recognize the emoticon and then compare
them and provide the result. This preprocessing does not give
any confusion and controversy in finding the opinion
B. Subjectivity Lexicon analysis:
Lexicon based techniques use a dictionary to perform
entity-level sentiment analysis. This technique uses
dictionaries of words annotated with their semantic orientation
(polarity and strength) and calculates a score for the polarity
of the document .It Uses r packages this r packages uses two
handy functions they are classify emotion and polarity. This
lexicon analysis is the analysis process of training the natural
language process into the machine learning process and the
meaning are oriented in the dictionary that which is said as the
trained. And it is compared and analyzed exactly whether the
particular context comes in under which emotion.
1. Classify emotions:
The emotions are classified into five categories such
as happy, unhappy, fear, disgust and joy. The each context in
the word are compared and analyzed with the trained dataset
and then the emotions are classified into the above five
categories. By using the naïve bays classifier trained on the
Carlo Straparva.
Class_emo=classify
emotion(some_txt,algorithm=“Bayes”,prior=1.0)
2. Classify polarity:
Classifying the polarity is the based on the context’s
emotion. The opinion is categorized into positive or negative
or neutral. And the percentage of the positivity or negativity or
neutrality will be provided from the live tweets for eg: if the
twitter dataset is of 5000 tweets the polarity of tweets would
be mentioned in the percentage level and the three colors are
used to differentiate.
Class_pol=classify_polarity(some_txt,algorithm=bayes)
If the above two process has to be done properly
mean the twitter authentication is important as developer of
Fig 2.1 Sentiment Analysis this project I got the twitter authentication and got the
A. Data Preprocessing and cleanup: authorization from the twitter to extract the live data that is
This module is responsible for preprocessing, clearing and live tweets from the many of the web users As the
tokenizing the original texts and converting them into authorization is provided the twitter package is installed in the
Document objects. r_console then the twitter access from the direct authentication
397
 International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353 Volume 22 Issue 2 – MAY 2016.

is and then the subjectivity analysis is processed and opinion other is the word cloud graph and the last one is aggregation
is extracted into three forms one is the polarity graph and the of the emotion and the context in the vertical line graph.

3. Process of extracting data and provide

Fig: 2.2 Process of providing the opinion

C. Emoticon Analyzer: D .Aggregation of emoticon and the context:

Emoticon analyzer is that encodes the UTF codes of This is the comparison in between the emoticon in
emoticons and makes recognize the emoticons in the context the context and the context itself. It compares each and every
and then it is analyzed as the positive or negative or neutral dataset with the emoticon’s polarity and the context’s polarity
emoticon. Here are the major 20 emoticons that which are and with their emotions and then aggregates them together and
majorly used by every common person in their day to day life. provides the overall opinion. Such as the positive or negative
And all those emoticon’s emotion are categorized and the or neutral and also provide the emotions categories positivity,
polarity in between the emoji’s are analyzed. Later it will be negativity and the Neutrality. So that user will be easily
compared and aggregated with context. deciding his/her opinion.

Table 1.1 Emoticon list

emoticons Notes
NO.
emoticons Notes
NO. #3 :( Sad face
#1 :) Happy face #4 ;) Wink
#2 :D Laugh Happy face
#5 :-)
(with nose)
398
 International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353 Volume 22 Issue 2 – MAY 2016.

emoticons Notes
NO.

#6 :P Tongue out

#7 =) Happy face
#8 (: Happy face
Wink (with
#9 ;-)
nose)
Uneasy,
#10 :/
undecided,
#11 XD Big grin
#12 =D Laugh
#13 :o Shock, Yawn
#14 =] Happy face
#15 D: Grin (mirror)
#16 ;D Wink and grin
#17 :] Happy face
Fig 3.1 Polarity Graph
#18 :-( Unhappy Polarity graph shows from overall tweets how many
Uneasy, were in positive and negative and neutral form.
#19 =/
undecided,

#20 =( unhappy

III. OVERALL RESULTS IN RGUI:

R is the interpreter language it executes the output
instantly .Here in this project the output executed in the R
Graphics Device (ACTIVE) That which is used to show the
execution in different graphical form.
This is the polarity graph, I searched the term
election from the twitter the tweets based on the twitter are
extracted from it and stored in the stack and the opinion scores
are given on the basis of naïve bays algorithm and the
sentiment packages.
More than 750 tweets were given only the positive
opinion about the election and the neutrality is more than 350
tweets and the negativity is below 350. So it is now useful for
the public to decide on the election.

Fig 3.2 Word Cloud

The word cloud is formed for the purpose of
detecting the emotion categories in the tweets.

399
 International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE)
ISSN: 0976-1353 Volume 22 Issue 2 – MAY 2016.

Fig 3.3 Aggregation of context and emoticon

[8]SentimentAnalysisforCantoneseOpinionMining, Jian Chen
IV. CONCLUSION School of Software Engineering South China University of
In this project I have increased the efficiency of Technology Guangzhou, China,2012
[9] Tweet Segmentation and Its Application to Named Entity
finding the opinion to 85% and in the existing system it was
Recognition Chenliang Li, Aixin Sun, JianshuWeng, and Qi
only up to the 60% and the accuracy of finding the exact He, Member, IEEE,2015
opinion was increased up to the 90% and I have made the [10]Polarity Consistency Checking for Domain Independent
emoticons to be recognized. And this project would be useful Sentiment Dictionaries Eduard C Dragut, Member, IEEE,
to many web users and make their decision easily without Hong Wang, Prasad Sistla, Clement Yu, and WeiyiMeng,
reading the number of reviews and feedback .And they do not Member, IEEE2015.
waste time in taking decision. BIOGRAPHY
V. FUTUREWORK 1. Priya C. completed B.Tech Information Technology in Agni
The future enhancement can improve the efficiency College of Technology and pursuing M.E. Computer Science
above the 90% and otheremoticons that were introduced and Engineering in Kingston engineering college, Vellore.
newly such as heart, nail polish, car, cheers etc.. Can be 2.SankarGanesh K M.E., M.I.S.T.E., Assistant professor,
analyzed.And provide more efficient analysis using R. Kingston Engineering College, Vellore., Completed B.E.
REFRENCES Computer Science and engineering in Anna University and
[1] Soman, K. P., DiwakarShyam and Ajay V. “Insight Into M.E. Computer Science and engineering in Annamalai
Data Mining: Theory And Practice”, PHI, 2009. University
[2] Jiawei Han, MichelineKamber , Jian Pei, “Data Mining:
Concepts and Techniques”, Third Edition (The Morgan
Kaufmann Series in Data Management Systems), 2012 [3]
David J. Hand, HeikkiMannila
andPadhraicSmyth“PrinciplesofDataMining”(AdaptiveCompu .
tationandMachineLearning),2005
[4]MininOpinion from Text Documents: A Survey Khairullah
Khan, Bah arum B.Baharudin, Aurangzeb Khan, Fazal-e-
Malik 2010.
[5] An Opinion-Tree based Flexible Opinion Mining Model
Juling Ding, Zhongjian Le, Ping Zhou, Gensheng Wang.

[6] An Approach Based on Tree Kernels for Opinion Mining

of Online Product Reviews Peng Jiang1, Chunxia Zhang2,
Hongping Fu1, Zhendong Niu1, Qing Yang1
[7] Opinion Mining and Social Networks: a Promising Match
Krzysztof Jędrzejewski, Mikołaj Morzy,2011
400