International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 2 (2016) pp 1404-1407
© Research India Publications. http://www.ripublication.com
1404
Analysis of real time social tweets for opinion mining
B. M. Bandgar
Research Scholar, Department of Computer Science,
Karpagam University, Coimbatore-641021, Tamil Nadu, India.
E-mail: bapuraob@yahoo.com
Dr. S. Sheeja
Associate Professor, Department of Computer Application,
Karpagam University, Coimbatore-641021, Tamil Nadu, India.
E-mail: sheejaajize@gmail.com
Abstract
We developed the indigenous Windows based user friendly
application in Java to extract, process and classify the real
time social network tweet using unstructured models. The
meaningful real time tweets are obtained and the same is used
for sentimental analysis. The processed meaningful tweets are
classified into three different opinion mining classes positive,
negative and neutral by using unstructured algorithms such as
EEC, IPC and SWNC model. The SWNC Model gave better
results over the EEC and IPC model. Their results are
compared using the confusion matrix, precision and accuracy
parameters. The results are also visualized using pie graph.
Keyword: Real time social tweets, Unstructured algorithms,
classification, opinion mining.
Introduction
The emergence of social media has given web users a venue
for expressing and sharing their thoughts and opinions on
different topics and events. Twitter, with nearly 600 million
users and over 250 million messages per day, has quickly
become a gold mine for organizations to monitor their
reputation and brands by extracting and analyzing the
sentiments of the tweets posted by the public about them, their
markets, and competitors. Sentiment analysis over Twitter
data and other similar micro-blogs faces several new
challenges due to the typical short length and irregular
structure of such content.
A number of researchers have carried out the items of the
related works. Cui, A. et al.[2] showed that sentiment analysis
of tweets is a challenging task due to multilingual and
informal messages. A. Bifet and E. Frank [3] proposed a data
mining technique used for sentiment knowledge in twitter data
streams. The proposed algorithm focuses on classification of
data streams and performs sentiment analysis in real time.
Only tweets containing an emoticon are considered; which is
very small sample of the tweet data.
A. Bifet, G. Holmes and B. Pfahringer[4] discussed the
handling of tweets in real-time. The authors use only positive/
negative classes for sentiment analysis. S. Argamon et al[5]
used a supervised learning algorithm for determining complex
sentiment-related attributes. These attributes were classified as
attitude type and force.
A. Nagy and J. Stamberger[6] proposed a technique to
efficiently identify the sentiment from disaster micro-blogs.
SentiWordNet 3.0 is used for sentiment detection from tweet.
The emoticon is also used as the pictorial representation of a
facial expression depicting the mood of a person as angry,
sad, normal or happy. Limitations of the technique are ability
to expand the initial seeds and continuous maintenance of
lists. A. Montejo-Raez et al.[7] proposed an unsupervised
approach for sentiment polarity detection from twitter tweets.
The limitations of the algorithms are handling of negation,
manual labeling process for certain tweets and facing flaws in
calculation of final polarity score. J. Kim et al.[8] have
proposed a collaborative filtering based model for the
prediction of sentiment in Twitter. They used Two Twitter
datasets for the evaluation of the proposed model. The results
showed the effectiveness of the proposed approach for
sentiment prediction. R. Machedon, W. Rand, and Y. Joshi [9]
classified social tweet messages into three categories;
informative, persuasive and transformative. The tweet data
was collected only for 65 music bands where each tweets
were labeled by human. The results showed that the proposed
method were effective only for the ‘informative’ category.
A. Balahur,[10] presented the sentiment analysis technique for
Twitter data. They have not given the comparison of results
with other.
Farhan Hassan Khan et. al.[11] tried to improve the accuracy
of text classification and resolve the data sparsity issues. They
proposed hybrid model Tweet opinion Mining (TOM) system.
They further processed the obtained tweets and used for the
classification using three different techniques. They classified
the tweets into positive, negative or neutral. This showed good
accuracy in the opinion mining. The limitations of their model
are no information of extraction of tweets and processing
method is not clear. The results obtained are given in
statistical terms only. They used the integrated system of
different tools in which integration is most complex and
developed system based on Linux operating system which is
not user friendly.
Therefore, we developed the user friendly indigenous
Windows based application in Java to extract real time tweets,
process and classify the opinions of the people all over the
global on the particular event and geographical area from the
different sources such as twitter website, news websites etc.
Method details
The twitter4j API version 1.01 downloaded from [12] and
created application on the web site [12]. The token and access