THE SURVEY OF SENTIMENT AND OPINION MINING FOR BEHAVIOR ANALYSIS OF SOCIAL MEDIA

International Journal of Computer Science & Engineering Survey (IJCSES) Vol.6, No.5, October 2015
DOI:10.5121/ijcses.2015.6502 21
THE SURVEY OF SENTIMENT AND OPINION MINING
FOR BEHAVIOR ANALYSIS OF SOCIAL MEDIA
Saqib Iqbal1
and Ali Zulqurnain1
, Yaqoob Wani1
and Khalid Hussain2
1
School of Information Technology, University of Lahore, Pakistan
2
Faculity of Computing, Universiti Teknologi Malaysia, Malaysia
ABSTRACT
Nowadays, internet has changed the world into a global village. Social Media has reduced the gaps among
the individuals. Previously communication was a time consuming and expensive task between the people.
Social Media has earned fame because it is a cheaper and faster communication provider. Besides, social
media has allowed us to reduce the gaps of physical distance, it also generates and preserves huge amount
of data. The data are very valuable and it presents association degree between people and their opinions.
The comprehensive analysis of the methods which are used on user behavior prediction is presented in this
paper. This comparison will provide a detailed information, pros and cons in the domain of sentiment and
opinion mining.
KEYWORDS
Sentiment Mining, Social Media Behaviour, Behaviour Prediction, Opinion Mining, Sentiments
1. INTRODUCTION
Social media has become an essential part of life, it is a comparatively cheap and widely
accessible medium that enables anyone to broadcast and access information / news / knowledge,
and to build new relations. It is a tool which is used to share different opinions that can belong to
different subjects; such as humanitarian causes, environmental irregularities, economic issues, or
political disputes. Honeycomb Framework describes social media by using seven functional
building blocks: identity, conversations, sharing, presence, relationships, reputation, and
groups[1].
Threaded deliberations, individuals not only express their views on an issue but also interact with
one another. However, the discussions could become extremely emotional and intense with many
expressive sentiments among individuals.
Microblogging (Twitter or Facebook) is the most popular communication channel among people
in this era. Many researchers are working on sentiment and opinion mining on textual information
which is gathered from the microblogging platforms. These sentiments and opinions are being
used to improve business’s marketing and productivity. Researchers employ the usage of several
mining techniques in order to extract the sentiments and to improve the sentiments extractions
results from textual, structured and un-structured formats.
Twitter (microblogging) revolution which occurred during Arab springs has raised two issues
which are now being discussed frequently i.e. the uses of Twitter (microblogging) for publicizing
local causes to distant audiences, and its importance in logistical communication among
protesters on the ground[2].

22
Although sentiment and opinion mining is very helpful in order to draw the behavior of users
against products, movies or events. More work is needed in order to grow the social media such
that it can allow individuals to monitor and predict any uncertainty, and how to address and cope
with such situations. Sentiment and opinion mining can become a more powerful tool by
improving and addressing how many users support a sentiment whether it is in negation or in
favor about a topic, and to identify the like-minded cluster of people and their relationship
through sentiment mining.
2. RELATED WORK
Nidhi Mishra et al., 2012has described the importance of opinion and sentiments mining. Opinion
mining is distributed into four sub-categories: sentence level, document level, feature level and
compound level respectively, and has discussed the various tools and mining methods like
precision, recall and F-measure on movie reviews or product reviews. Aspect level classification
can provide greater detail for analysis then document and sentence level. The distribution
categories miss classification subjective expressions and objective expressions for better
understanding of sentiments [3,4,5].
Meeyoung Cha et al., 2010investigated the dynamics of public influence across different issues /
topics to determine the influence author used in degree, retweets and mentions approaches on a
large dataset of twitter. Spearman's rank correlation coefficient has been used for evaluating the
influence of people against three different topics. Author concludes the result in terms of
influence against each topic but he didn’t target the fake users which can actively participate in
discussions. Fake user can play the major role to divert the influence of public from one issue to
another. [6,7].
Guendalinaet al., 2015presents a new approach known as the integrated mixed methods approach
(IMiME) for monitoring the content of social media by demonstrating a case study ‘hypoactive
sexual desire disorder (HSDD). Integrated Mixed methods approach (IMiME) is the combination
of qualitative and quantitative approaches. Although author has presented the result of IMiME
approach as balanced between ecological and situated nature of qualitative research strategies
with systematization of quantitative measurement. This approach have to be tested with other
case studies for concrete satisfaction [8, 9, 10].
Rui Gaspar et al., 2014has assessed the ways of coping, expressed by public in social media
(Twitter) during food crises for its management and prevention. Methodological, conceptual
limitation and challenges are discussed for coping analysis and how to apply psychosocial
process to cope the content of single tweet. Qualitative approach is used for cope analysis.
Qualitative approach is much hard and time consuming to analyze large dataset. Author selected
limited data after cleansing for better results. During selection of data, author didn’t consider the
spam opinions. Spam opinions can be more harmful for decision making. Qualitative approach is
used, if the Quantitative approach may also be added than it will be helpful for coping analysis
and supportive for decision making [11, 12, 13, 14].
JianshuWenget al., 2010 explored the most influenced users of Twitter (micro-blogging site) by
homophily. In homophily, a single user choose and follow a friend on twitter because he is
interested in the topics which were published in tweets by the friend and friend follows back
because he observes that they both have similar interest topics. Author endorse the homophily is
not being used in case of twitter by using TwitterRank Approach. TwitterRank approach is the
combination of Latent Dirichlet Allocation (LDA) and TwitterRank algorithms. Topic-sensitive
PageRank has also been discussed, and its comparison with TwitterRank has been put forth. A
group of spam/fake users can add someone to become more influenced user like celebrities,
politicians etc. The author didn’t detect the group of fake users as optimistic approach [15, 16].

23
Federico Neriet al., 2012has applied the sentiment mining analysis on Facebook textual dataset
for news channel by using Recall and Precision approach. In this approach, it is very difficult to
identify what is relevant to one person might not be relevant to another person and vice versa.
Naïve Bayes can be used instead of Precision and Recall for better results [17, 18].(
Jayasankaet al., 2014has assessed the challenges which were observed during extraction of
expressions of optimistic or pessimistic attitudes through social media and has also discussed the
different available tools and their limitations for sentiment analysis. Naïve Bayes classification
method which is used in order to produce better results of sentiment mining instead of Natural
language Processing and SentiWordNet. Scores are used in order to project the results of negative
and positive sentiments. Expectation-Maximization (EM) algorithm can also be used in order to
optimize the accuracy as the SentiWordNet does not make much difference among short
sentences like ‘not good’ and ‘not bad’ [19, 20]. (
Ion SMEUREANU et al., 2012has presented a work for exploring the positive and negative
opinions on pre-classified movies’ reviews. Naïve Bayes Algorithm was applied on collection of
comments. Precision and recall methods are used for accuracy check of Naïve Bayes algorithm
and its execution time [21, 22].
Tkalcicet al., 2014has presented the scenario and design of personalized intervention system
(PIS) and has chosen theory of planned behavior (TPB) method to predict whether user will
attend the classical musical concert or not through social media mining. As theory of planned
behavior (TPB) does not consider the emotional elements like feelings, mode; emotion element
which has direct effect on personalized intervention system (PIS) predictions is not
considered[23].
Sergio Oramas., 2014has used the social media (micro-blogging sites) and web (Musical content
sites) for Musical Information Retrieval (MIR). Social Media Mining, Knowledge Extraction and
Natural Language Processing techniques are combined in Musical Information Retrieval (MIR).
Musical Information Retrieval (MIR) will work on unstructured and structured data. Structured
and unstructured data will get from web and micro-blogging sites respectively[24].
Ahmed Hassan et al., 2010has presented a method in order to detect an individual’s attitude in a
sentence towards others in threaded base communication. This approach is a combination of
supervised Markov model of text, part-of-speech and dependency patterns. Support Vector
Machines (SVM) is used for experiment results and to compare with other baselines: which are
accuracy, recall and precision and F1[25].
Robert McColl et al., 2014has performed a comparison of execution and performance with
twelve(12) open source graph databases. The Author has chosen same set of hardware, nodes and
edges for execution and performance. Each open source graph database is evaluated with each
algorithm which are Single Source Shortest Path (SSSP), Connected Components, PageRank and
Update Rate[26].
Praveshet al., 2014has assessed and evaluated the numerous methods which are used for opinion
and sentiment. Naïve Bays Classifier, Support Vector Machine (SVM), Multilayer Perceptron,
Clustering techniques has been used in order to analyze and compare the results of each
technique. Every technique offer various benefits and boundaries, each method can be utilized
according to the situation for features and text extractions. There are many other methods which
can be utilized for extraction of sentiments from micro-blogging sites[27].

24
Anindya Ghose et al.,2007 has proposed two ranking mechanisms: consumer-oriented ranking,
manufacturer-oriented mechanism. These mechanisms are consisted on econometric and
subjectively analysis and text mining. Consumer-oriented mechanism is used in order to rank the
reviews according to their expected helpfulness while manufacturer-oriented mechanism is used
in order to rank the reviews according to their expected effect on sales. Reviews can be spam or
genuine but in these mechanism no spam detection method has been used. If spam reviews are
also included in the ranking system, it will spoil the results and detract the community [28].
Table-1: Comprehensive research Analysis of Sentiment and Opinion Mining
Paper
Ref.
No
Techniques/Methods/Protocols Advantages Disadvantages Remarks
[3]
1. Document Level
2. Sentence level
3. Task of Opinion mining at
Feature level
4. Opinion Mining in
Compound sentence
Author classify the
opinion into different
levels for
understanding and
mining.
Document and
sentence level
mining does not
provide the
necessary detail
needed for many
other applications.
Aspect level
Opinion mining
should also
include in
classification
for better
understand
[6] Spearman’s rank coefficient
Author use
quantitative approach
and project the users
influence against
different
subjects/objects
Didn't detect the
Fake users and
follows
Fake users can
change the
results
[8]
Integrated mixed
Methods(IMiMe) approach
Mixed-methods
approach is
combination of
different qualitative
and quantitative
analytical strategies to
overcome the
drawbacks of available
methods/tools.
IMiMe approach is
case specific and
needed to be
checked and
verified with other
cases
Understanding
and prevention
User behavior
[11]
Coping Analysis/ psychosocial
processes
Determine Ways of
coping from social
media and discuss the
methodological,
conceptual limitations
and challenges
Spam opinion
detection is
missing, which can
become the cause
of worse decision
Opinion Spam
Detection still a
challenging
problem which
need to be sort
out
[15]
Latent Dirichlet Allocation
(LDA) /TwitterRank
Author used
Unsupervised learning
and quantitative
approach for track the
most influencer users
of Twitters
Group spammers
can play a the
major role to make
someone the most
influencer user,
The detection of
group spammer is
missing
Group
spammers can
mislead the
community and
change the
thought.
[17] Recall and Precision
Author used Social
media specifically
Facebook for
sentiment mining for
marketing
Precision and
Recall, difficult to
identify what is
relevant or
irrelevant to one
person may not be
relevant/irrelevant
to another
Naïve Bayes
can be used
instead
Precision and
Recall for better
results.

25
[18]
Naïve Bayes Classification
,Maximum Entropy, Natural
Language Tool Kit (NLTK),
SentiWordNet
Author discuss
sentiment extraction
challenges and apply
different techniques
on data.
SentiWordNet
cannot analyze
short terms words
like 'not bad' or 'not
good' which reduce
the accuracy
Expectation-
Maximization
(EM) algorithm
can use to
optimize the
accuracy
[21] Naive Bayesian classifier
Author verify the
efficiency of
Algorithm on positive
and negative on user
reviews.
Naïve Bayes can’t
learn interactions
between features
Accuracy of
SVM is better
than Naïve
Bayes, must try
SVM instead
Naïve Bayes
[23]
Theory of Planned Behavior
Model (TPB)
TPB model is used for
predicting the
behavior of users
theory of planned
behavior overlooks
emotional
variables such as
threat, fear, mood
and negative or
positive feeling
Theory of
planned
behavior is not
useful to
prediction
because it
doesn't consider
emotions
[24]
Music Information Retrieval
(MIR).
a methodology that
combines Social
Media Mining,
Knowledge Extraction
and Natural Language
Processing techniques,
to extract meaningful
context information
for music from social
data.
Application will
not distinguish
between
comparative
sentiments and
regular sentiments.
should also
include
comparative and
regular
sentences
mining for
accuracy in
mining the
textual
[27]
1. Single Source Shortest Path
(SSSP)
2. Connected Components
3. PageRank
4. Update Rate
Discuss the execution
power and comparison
between open Source
Graph databases with
same hardware and
collection of edges.
Support vector
machine can also
be utilized for
comparison
between open
source graph
databases
Open Source
Graph Database
can help to
analyses the
like-minded
group of people.
[25]
Markov model of text, part-of-
speech, dependency patterns
it predicts whether a
sentence contains an
attitude toward a text
recipient or not
Not all subjective
sentences have
opinions
Can implement
on
microblogging
sites for analysis
[26]
1. Naïve Bays Classifier
2. Support Vector Machine
(SVM)
3. Multilayer Perceptron
4. Clustering
Author evaluate
different algorithms of
Opinion and sentiment
mining and compare
them
Precision and
Recall technique
can also compare
for sentiment and
opinion mining
User's
comments can
be used to draw
like-minded
people and their
interests.
[28]
consumer-oriented ranking,
manufacturer-oriented ranking
ranking mechanism is
combination of
econometric analysis,
text mining and
subjectivity analysis
Reviews can be
Spam or not spam,
it really effect on
market for selling
product
Spam reviews
should not be
included during
ranking of any
product
3. DISCUSSION
Sentiment and opinion mining is a very crucial / ambiguous process for extraction and evaluation
of the behavior of public sentiment from social media. In this paper we discussed different
research papers regarding sentiment analysis and opinion mining presented in Table-1.
Researchers utilize and evaluate different algorithms and approaches forsentiment mining on

26
different subjects, different areas of sentiment and opinion mining are targeted to improve mining
process. This research-works is very beneficial for improving the analysis and behavior
prediction.
Prime focus of this study is classification of sentiment, evaluate the efficiency and accuracy of
different sentiment and opinion mining algorithms like SVM (Support Vector Machine), Naïve
Bayes Classification, Latent Dirichlet Allocation (LDA) , PageRank, Markov model of text,
Multilayer Perceptron and Clustering.
Social media is a massive source of communication, where every individual post their sentiments.
Detection of bogus or fraudulent sentiments or reviews become very hard from these sentiments
and reviews which can post through fake users and spammers. These bogus or fraudulent
sentiment can become a powerful approach and helpful to divert the mindset of individuals to
boost any product demand and market capturing. Social media support many native/national
languages of different countries which make hard to analyze sentiment of different countries on
same topic or event. Use of informal or casual language reviews or opinions is another major
issue during sentiment analysis because understanding of causal or informal language is much
difficult then the formal language.
4. CONCLUSION
In Social media, people usually follow various events such as natural disasters, political issues,
sports events and posts there comments / sentiments about the particular event. By applying
sentiment and opinion mining techniques, it can be deduced that whether the sentiment is in
denial or in favor of the event. But the most important point is how many people in total support
and broadcast this particular sentiment, and how to identify the relationship amongst the
supporters of a specific sentiment. This approach will be able to predict any uncertainty which
may be either harmful or beneficial for the institutions.
A detailed assessment of different papers / articles is presented in this paper. There are several
work which has been done on sentiment and opinion mining of social media but this mining
technique is currently unformed. There are various research gaps in sentiment and opinion mining
which are explained in the work done by the researchers. Many researchers are still researching
on sentiment and opinion mining and how to improve this particular mining techniques.This
paper is based on a comprehensive study on sentiment and opinion mining, and discuss the
limitations and address the new areas in the domain of sentiment and opinion mining.
REFERENCES
[1] Kietzmann, J. H. (2011). Social media? Get serious! Understanding the functional building blocks of
social media. Business horizons , 54- 3, 241-251.
[2] Segerberg, A. a. (2011). Social media and the organization of collective action: Using Twitter to
explore the ecologies of two climate change protests. The Communication Review , 14-3, 197-215.
[3] Mishra, N., & C.K.Jha. (2012). Classification of Opinion Mining techniques. International Journal Of
Computer Applications, 0975-8887.
[4] Jaganadh, G. (2012). Opinion mining and Sentiment analysis CSI Communication.
[5] Liu, B. (2011). Opinion Mining and Sentiment Analysis. AAAI, San Francisco,.
[6] Cha, M., Haddadi, H., Benevenuto, F., & Gummadi, K. P. (2010). Measuring User Influence in
Twitter: The Million Follower Fallacy. ICWSM, 10, 10-17.
[7] Goyal, A., Bonchi, F., & and Lakshmanan, L. V. (2010.). Learning Influence Probabilities In Social
Networks. ACM WSDM.
[8] Graffigna, G., & Riva, G. (2015). Social Media Monitoring and understanding : an integrated mixed
methods approach for the analysis of social media. International Journal of Web Based Communities,
11.1, 57-72.

27
[9] Laine, M. a. (2010). Monitoring social media: tools, characteristics and implications. Springer Berlin
Heidelberg., 51- 2, 193–198.
[10] Ellonen, H.-K. A. (2010). The effect of website usage and virtual community participation on brand
relationships. International Journal of Internet Marketing and Advertising, 56-1, 85-105.
[11] Gaspar, R. S. (2014). Tweeting during food crises: A psychosocial analysis of threat coping
expressions in Spain, during the 2011 European EHEC outbreak. International Journal of Human-
Computer Studies, 72.2 , 239-254.
[12] Newkirk, R. W. (2012). The potential capability of social media as a component of food safety and
food terrorism surveillance systems. Foodborne pathogens and disease, 9-2, 120-124.
[13] Sheth, A. P. (2010). Understanding events through analysis of social media.
[14] Thelwall, M. K. (2011). Sentiment in Twitter events. Journal of the American Society for Information
Science and Technology, 62-2, 406-418.
[15] Weng, J. E.-P. (2010). Twitterrank: finding topic-sensitive influential twitterers. In Proceedings of the
third ACM international conference on Web search and data mining,, 261-270.
[16] Milstein, S. B. (2008). Twitter and the micro-messaging revolution: Communication, connections,
and immediacy. 140 characters at a time. O'Reilly Media, Incorporated,.
[17] Federico, N. C. (2012). Sentiment analysis on social media. IEEE/ACM International Conference on
Advances in Social Networks Analysis and Mining.
[18] Pang, B. a. (2008). Opinion mining and sentiment analysis. Foundations and trends in information
retrieval, 2(1-2), 1-135.
[19] R. A. S. C. Jayasanka, M. D. (2014). Sentiment Analysis for Social Media.
[20] Osimo, D. a. (2012). Research challenge on opinion mining and sentiment analysis. Universite de
Paris-Sud, Laboratoire LIMSI-CNRS, Bâtiment , 508.
[21] Smeureanu, I. a. (2012). Applying Supervised Opinion Mining Techniques on Online User Reviews.
Informatica Economică, 16-2.
[22] Wang, H. a. (2008)). A knowledge management approach to data mining process for business
intelligence. Industrial Management & Data Systems, 108-5, 622-634.
[23] Tkalcic, M. B. (2014). Using social media mining for estimating theory of planned behaviour
parameters.
[24] Oramas, S. (2014.). Harvesting and structuring social data in music information retrieval. In The
Semantic Web: Trends and Challenges,Springer International Publishing, , 817-826.
[25] Hassan, A. V. (2010). What's with the attitude?: identifying sentences with attitude in online
discussions. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language
Processing, (pp. 1245-1255).
[26] McColl, R. C. (2014). A performance evaluation of open source graph databases. ACM(In
Proceedings of the first workshop on Parallel programming for analytics applications), 11-18.
[27] Singh, P. K. (2014). Methodological study of opinion mining and sentiment analysis techniques.
International Journal on Soft Computing, 5(1), 11.
[28] Ghose, A. a. (2007). Designing novel review ranking systems: predicting the usefulness and impact of
reviews. In Proceedings of the ninth international conference on Electronic commerce. ACM.

THE SURVEY OF SENTIMENT AND OPINION MINING FOR BEHAVIOR ANALYSIS OF SOCIAL MEDIA

More Related Content

What's hot (19)

Viewers also liked (17)

Similar to THE SURVEY OF SENTIMENT AND OPINION MINING FOR BEHAVIOR ANALYSIS OF SOCIAL MEDIA (20)

Recently uploaded (20)

THE SURVEY OF SENTIMENT AND OPINION MINING FOR BEHAVIOR ANALYSIS OF SOCIAL MEDIA