International Journal of Computer Applications (0975 – 8887)

Volume 56– No.13, October 2012

Classification of Opinion Mining Techniques

Nidhi Mishra C.K.Jha, PhD.

Assistant Professor, Department of Information Associate Professor, Department of AIM & ACT,
Technology, Buddha Institute of Technology Banasthali Vidyapith, Rajasthan India
Gorakhpur, India

ABSTRACT products and services. The introduction of Blog track in

The important part to gather the information is always seems TREC 2006, a considerable work has been done in this field
as what the people think. The growing availability of opinion which comprises of opinion mining at sentence level, passage
rich resources like online review sites and blogs arises as or document level and feature level. There are many
people can easily seek out and understand the opinions of challenges in opinion mining. The first challenges is that
others. Users express their views and opinions regarding opinion word is considered to be positive in one condition
products and services. These opinions are subjective may be considered negative in another condition. A second
information which represents user’s sentiments, feelings or challenge is that people do not always state the opinions in a
appraisal related to the same. The concept of opinion is very same way. The art Opinion Mining is to recognize the
broad. In this paper we focus on the Classification of opinion subjectivity and objectivity of a text and further classify the
mining techniques that conveys user’s opinion i.e. positive or opinion orientation of subjective text. In short we say that
negative at various levels. The precise method for predicting Opinion Mining[19] is an automated extraction of subjective
opinions enable us, to extract sentiments from the web and content from text and identifying the orientation such as
foretell online customer’s preferences, which could prove positive or negative in that text. It aims to explore feelings of
valuable for marketing research. Much of the research work a person who write the text. It used Natural Language
had been done on the processing of opinions or sentiments Processing and Machine Learning ethics to determine opinion
recently because opinions are so important that whenever we in the text. The evaluation of opinion can be done in two
need to make a decision we want to know others’ opinions. ways:
This opinion is not only important for a user but is also useful  Direct opinion, gives positive or negative opinion
for an organization. about the object directly. For example, “The picture
quality of this camera is poor” expresses a direct
Keywords opinion.
Opinion Mining; Machine learning; Sentiments; Polarity;
semantic.  Comparison means to compare the object with some
other similar objects. For example, “The picture
1. INTRODUCTION quality of camera-y is better than that of Camera-x.”
We Opinion mining is a type of natural language processing expresses a comparison.
for tracking the attitudes, feelings or appraisal of the public
about particular topic, product or services. Textual Spam filtering refers to detection and removal of fake
information in the entire world is of two types: facts and opinions that mislead the users by giving unworthy positive or
opinions. The facts are the objective expressions which negative opinions to some objects in order to sponsor or spoil
describe the entities, events and properties whereas the the objects reputations. It is also a research issue in healthy
opinion is the subjective expression which describes people's opinion mining. The applications of opinion mining are in (1)
opinions, emotions and sentiments towards entities and their Argument mapping software helps organizing in a logical way
properties. The current search engine searches for facts these policy statements, by explicitating the logical links
because they assume the facts are true and can be expressed between them (2) Voting Advise Applications: help voters to
with keywords. But these search engines do not find the understand which political party have closer positions with
opinions because opinions or sentiments are very difficult to them. (3) Automated content analysis helps processing large
express by keywords and that is why there ranking strategy amount of qualitative data. There are lots of Free and Open
are not appropriate for opinion retrieval. Now, the web has Source tools available for performing Natural Language
significantly changed its way that people comments their Processing and Machine Learning tasks. Also the huge
views and opinion on any product and services. User can post amount of user generated content, organized for opinion
their comments on any internet forums, review sites, blogs mining task is available on web. Following tools like GATE,
and discussion group which are commonly known as user NLTK, Apache Mahout, Weka, Rapidminer, KNIME, and
generated content which contains the important information. Open NLP etc. can be used to develop your own opinion
This online word- of- mouth behavior represents new and mining automated system. This paper is an extension to our
considerable sources of information and their applications. work is which is carried out in my paper [28] titled- “An
These online comments are not limited to our friend circle but insight into task of opinion mining” accepted in Springer
it is also expanded on a global or web scale. Today, if the user SPIT-2012 International Conference.
wants the views on a particular product he/she has no longer Rest of the paper is organized as follows Section 2 deals about
limited to ask their friends because they got the opinion for literature review of classification of opinion mining
that product on the internet through various reviews or techniques and opinion mining in compound sentence.
comments. In the same if the organization wants the opinion Section 3 deals the discussion and comparison of various
of their products and services they use these user generated opinion mining techniques. In section 4 discuss about various
contents from the web for the review or comments of their

1
 International Journal of Computer Applications (0975 – 8887)
Volume 56– No.13, October 2012

opinion mining tools. Finally we conclude our discussion in 2.2 Task of opinion mining at Sentence level
Section 5. The sentence level opinion mining is associated with two
2. LITERATURE REVIEW AND TASK tasks [6] [7] [8]. First one is to identify whether the given
sentence is subjective (opinionated) or objective. The second
OF OPINION MINING one is to find opinion of an opinionated sentence as positive,
All In order to give more imminent into the problem of
negative or neutral. The assumption is taken at sentence level
opinion mining, in the following sections we discuss the
is that a sentence contain only one opinion for e.g., “The
domain overview and various types of opinion mining. The
picture quality of this camera is good.” However, it is not true
opinion mining is frequently associated with the topic
in many cases like if we consider compound sentence for e.g.,
information retrieval. The information retrieval algorithm
“The picture quality of this camera is amazing and so is the
works on factual data but the opinion mining works on
battery life, but the viewfinder is too small for such a great
subjective data. The task of opinion mining is to find the
camera”, expresses both positive and negative opinions and
opinion of an object whether it is positive or negative and
we say it is a mixed opinion. For “picture quality” and
what features does it depict, and what features are appreciated,
“battery life”, the sentence is positive, but for “viewfinder”, it
which are not etc. The notion of an opinion mining is given by
is negative. It is also positive for the camera as a whole. Riloff
Hu and Liu [2]. They put most impact on their work and said
and Wiebe [11] use a method called bootstrap approach to
that the basic components of an opinion are:
identify the subjective sentences and achieve the result around
 Opinion holder: it is the person that gives a 90% accuracy during their tests. In contrast, Yu and
specific opinion on an object. Hatzivassiloglou [13] talk about sentence classification
 Object: it is entity on which an opinion is expressed (subjective/objective) and orientation
by user. (positive/negative/neutral). For the sentence classification,
 Opinion: it is a view, sentiment, or appraisal of an author’s present three different algorithms: (1) sentence
object done by user. similarity detection, (2) naïve Bayens classification and (3)
multiple naïve Bayens classification. For opinion orientation
authors use a technique similar to the one used by Turney [27]
2.1 Task of Opinion Mining at Document for document level. Wilson et al. [12] pointed out that not
level only a single sentence may contain multiple opinions, but they
Document level opinion mining is about classifying the also have both subjective and factual clauses. It is useful to
overall opinion presented by the authors in the entire pinpoint such clauses. It is also important to identify the
document as positive, negative or neutral about a certain strength of opinions. Like the document-level opinion mining,
object [3] [4]. The assumption is taken at document level is the sentence-level opinion mining does not consider about
that each document focus on single object and contains object features that have been commented in a sentence. For
opinion from a single opinion holder. Turney [27] present a this the feature level opinion mining is discuss in the next sub-
work based on distance measure of adjectives found in whole section.
document with known polarity i.e. excellent or poor. The
author presents a three step algorithm i.e. in the first step; the 2.3 Task of Opinion mining at Feature level
adjectives are extracted along with a word that provides The task of opinion mining at feature level is to extracting the
appropriate information. Second step, the semantic orientation features of the commented object and after that determine the
is captured by measuring the distance from words of known opinion of the object i.e. positive or negative and then group
polarity. Third step, the algorithm counts the average semantic the feature synonyms and produce the summary report. Liu
orientation for all word pairs and classifies a review as [16] used supervised pattern learning method to extract the
recommended or not. In contrast, Pang et al. [5] present a object features for identification of opinion orientation. To
work based on classic topic classification techniques. The identify the orientation of opinion he used lexicon based
proposed approach aims to test whether a selected group of approach. This approach basically uses opinion words and
machine learning algorithms can produce good result when phrase in a sentence to determine the opinion. The working of
opinion mining is perceived as document level, associated lexicon based approach [18] is described in following steps.
with two topics: positive and negative. He present the results
using nave bayes, maximum entropy and support vector  Identification of opinion words
machine algorithms and shown the good results as comparable  Role of Negation words
to other ranging from 71 to 85% depending on the method and  But-clauses
test data sets. Apart from the document-level opinion mining,
the next sub-section discusses the classification at the
sentence-level, which classify each sentence as a subjective or
objective sentence and determine the positive or negative
opinion.

2
 International Journal of Computer Applications (0975 – 8887)
Volume 56– No.13, October 2012

Table 1. Presents insight into opinion mining at different In contrast , Hu and Liu do customer review analysis [26]
levels[28] through opinion mining based on feature frequency, in which
Classification Assumptions made at Tasks associated the most frequent features is accepted by processing many
of Opinion different levels with different reviews that are taken during summary generation. In opposite
mining at levels to Hu and Liu, Popescu and Etzioni [20], improved the
different frequency based approach by introducing the part-of
levels relationship and remove the frequent occurring of noun
phrases that may not be features.
1. Opinion 1. A sentence Task 1:
Mining at contains only one identifying the
Sentence opinion posted by 2.4 Opinion Mining in Compound sentence
given sentence as
level. single opinion In this sub-section the following methodology we use to
subjective or determine the opinion in compound sentence of a movie
holder; this could
not be true in opinionated review domain:
many cases e.g. Classes:
there could be objective and 2.4.1 Sentence classification
multiple opinions subjective In the sentence classification we go to individual compound
in compound and sentences to determine whether a sentence is subjective or
(opinionated) express an opinion and if so, whether the opinion is positive
complex Task 2: opinion
sentences. or negative (called sentence-level sentiment classification).
classification of For example, 'Desi Boyz' - highly entertaining comedy
2. Secondly the
sentence the given gives the positive opinion and ‘Damadamm’ clearly has no
boundary is sentence. Dum gives the negative opinion.
defined in the Classes: positive, The following activities are done within sentence
given document classification:
negative and
neutral.
2. Opinion 1. Each document Task 1: opinion
2.4.1.1 Splitting of the document into sentences
Mining at focuses on a classification of Given a document about the movie reviews, the document is
Document single object and reviews segmented into individual sentences by the help of sentence
level. contains opinion Classes: positive, delimiter. Here problem is that most of the reviews are found
posted by a single negative, and on movie forums or blog sites where normal users post their
opinion holder. neutral opinions in their informal language which do not follow strict
2. Not applicable grammatical rules and punctuations. The identification of full
for blog and stop in the sentence does not mark the end of sentence
forum post as sometimes. Such as date 12.1.2012, movie short forms T.M.K
there could be expressed in example 1, hence we have to use rule based
multiple opinions pattern matching to identify sentence boundary. Second
on multiple problem is that people generally use internet slang words like
objects in such OMG, cuteeeee etc. e.g. actress is cuteeee. Here there is not
sources. such word in dictionary like cuteeee but it refers to cute. We
will do N gram matching of such words with pre compiled
3. Opinion 1. The data source Task 1: Identify
dictionary of movie related words. The output of this splitting
Mining at focuses on and extract object
document into sentences yields following:
Feature features of a features that have
level. single object been commented  M Gud reviews about film released on 12.1.2012.
posted by single on by an opinion  He says, “The film T.M.K’s story is filled with a
opinion holder. holder (e.g., a great plot, the actors are first grade, and actress is
2. Not applicable reviewer). cuteeee”.
for blog and Task 2:  The supporting cast is good as well, but, movie
forum post as Determine can’t hold up.
there could be whether the
multiple opinions opinions on the 2.4.1.2 Determining whether the sentence is
on multiple features are opinionated
objects in such positive, negative We will use boot strap approach proposed by Riloff and
sources. or neutral. Wiebe [11] for the task of subjective sentences identification.
Task 3: Group It will use a high precision (and low recall) classifiers to
feature extract a number of subjective sentences collected from
synonyms. various movie review sites. From this subjective sentence a
Produce a set of patterns will be learned. The learned patterns will be
feature-based used to extract more subjective and objective sentence. The
opinion summary subjective classifier will look for the presence of words from
of multiple the pre compiled list, while the objective classifier tries to
reviews. locate sentences without those words. In the example 1 all
sentences except “M Gud reviews about film released on
12.1.2012” are opinionated.

3
 International Journal of Computer Applications (0975 – 8887)
Volume 56– No.13, October 2012

2.4.1.3 Determining whether the opinionated

sentence is compound sentence 3. DISCUSSION AND COMPARISON
A compound sentence is a sentence that contains two or more
The performance of opinion mining in determining the
complete ideas (called clauses) that are related. These two or
orientations or polarity is evaluated by calculating various
more clauses are usually connected in a compound sentence
metrics like precision, recall and F-measure. The overview of
by a conjunction. The coordinating conjunctions are "and",
the work done in the task of opinion mining and its
"but", "for", "or”, “nor", "yet", or "so". We will use plain
orientations is discussed and compare in the Table II [17]. It is
pattern matching to find out the presence of coordinating
seen from the Table II, as the data source is concerned, a huge
conjunctions. If they are present in the given sentence then it
amount of work has been done on movie and product reviews
will be identified as compound sentence. Output of this will
to determine the opinion orientations. The Internet Movie
be set of compound sentences
Database is used for movie reviews and product reviews are
 He says, “The film T.M.K’s story is filled with a
taken from Amazon.com. Movie review is a more challenging
great plot, the actors are first grade, and actress is application than many other types of review mining. The
cuteeee”. challenges of movie review based on the factual information
which is always mixed with real-life data and mocking words
The supporting cast is good as well, but, movie can’t hold up
are used in writing movie reviews. The Product review
The compound sentence is a sentence which expresses domain considerably differs from movie review domain
more than one opinion of an object. For example, the because of the following reasons. One reason is that there are
sentence, “The supporting cast is good as well, but, movie
feature specific comments in product reviews because people
can’t hold up.”, expresses both positive and negative opinions
may like some features and dislike others. Thus reviews
(one may say that it has a mixed opinion). For “supporting consists opinions orientations in the text, which is a difficult
cast”, the sentence is positive, but for “overall”, it is negative. one to classifying opinion orientation of review as positive or
It is also positive for the movie as a whole.
negative. Following feature specific reviews occur less often
in movie reviews. Second reason is that there are a lot of
2.4.2 Opinion Mining in Compound sentence comparative sentences in product reviews and people discuss
In this section we focus on opinion expressions in the about other products in reviews.
compound sentence of a movie review that gives the opinion
on the individual feature of the movie and the opinion of the 4. TOOLS USED IN OPINION MINING
movie as whole i.e. positive or negative sentiments. Apart The tools which are used to track the opinion or polarity from
from this we also determine the sentiment score towards the user generated contents are:
various features of a movie, such as cast, director, story
and music. Sentiment scores are used to classify the  Review Seer tool [17] – This tool is used to automates
sentiment polarity (i.e. Positive, negative or neutral) of the work done by aggregation sites. The Naive Bayes
clauses or sentences. The linguist approach makes use of both classifier approach is used to collect positive and
a domain-specific lexicon (specify the noun related terms like negative opinions for assigning a score to the extracted
actor, director etc.) and a generic opinion lexicon(specify feature terms.
the property of movie related terms),derived from  Web Fountain [25]- It uses the beginning definite Base
SentiWordNet[9] ,to assign a prior sentiment score to Noun Phrase (bBNP) heuristic approach for extracting
each word in a sentence. For example, the product features.
 He says, “The film T.M.K’s story is filled with a  Red Opal [24] –It is a tool that enables the users to
great plot, the actors are first grade, and actress is determine the opinion orientations of products based on
cuteeee. their features. It assign the scores to each product based
 The supporting cast is good as well, but, movie on features extracted from the customer reviews’
can’t hold up.  Opinion observer[17]-This is an opinion mining system
The compound sentences (1 and 2) are divided into “The for analyzing and comparing opinions on the Internet
film T.M.K’s story is filled with a great plot” “the actors are using user generated contents. This system shows the
first grade” “actress is cute”; “The supporting cast is good as results in a graph format showing opinion of the product
well” and “movie can’t hold up”. After dividing the feature by feature.
compound sentences into separate sentences or clauses, a Along with these automated tools, there are various online
sentiment score toward each movie features (e.g. Story, tools like Twitrratr, Twendz, Social mention, and Sentimetrics
actors, actress, supporting cast or the film features) is are available to track the opinions in the web.
calculated. (For example like the sentiment score is from 0 to
5 here 0 indicates the most negative opinion ,5 indicate the
most positive opinion and between 0 to 5 we have to make the
rules for positive opinion and negative opinion.) After
calculating the sentiment score for each clause or sentence,
the sentiment score for each review features and the
overall sentiment score for the whole sentence is
calculated. The following tools we use like: Sentence
Delimiter [15], Part of Speech tagger [14], Named Entity
recognition [14], Sentiword Net [9] for determine the opinion
in compound sentence.

4
 International Journal of Computer Applications (0975 – 8887)
Volume 56– No.13, October 2012
Table 2. Summary based on literature survey
of opinion mining[17]

Performan
Technique

Precision
selection
5. CONCLUSIONS

Feature
Studies

Mining

Source

Recall
This paper presents the classification of opinion mining

Data
used
S.no

F1
techniques. Opinion mining aims at recognizing, classifying

ce
and determining opinion orientations of the opinionated text.
In this paper we first presented a theoretical model of opinion
Multiclass

Linguistic
Xu(2011)

mining, which discuss the problem and supply a common

Kaiquan

Amazon
Review
Feature

61.9%

93.4%

74.2%
outline in different research directions. It then discussed in
SVM

61%
section 2 the most usually topic task of opinion mining
1

techniques at various level, which determines whether a

document or sentence is carries a positive or negative opinion.
dependency

ME-85.4%
NB-85.8% We also discuss the feature level opinion mining. We observe
Xia(2011)

maximum
Unigram,

grammar

Amazon
entropy,
Bigram,

review,

that the opinion mining in compound sentence is also very

Bayes,

Movie

86.4%
SVM-
MDS,
Naïve

SVM

challenging task which we already discuss in section 2. In

Rui

section 3 we discuss about comparison of various opinion

2

-
mining techniques. Last but not the least we discuss about the
Dependanc

opinion spams in text. Detecting fake opinions is a research

Maximum
somprasti

y relation
Gamgam

Amazon

issue. Some of the opinions are time sensitive hence

entropy

review
(2010)

72.6%

78.7%

75.4%
identifying latest opinions on the subject is another research
issue. In section 4 we discuss about various opinion mining
3

tools used to track the opinion in the web. Finally, we

conclude the paper by proverb that all the tasks of opinion
clustering

mining and detecting the opinion orientation are very

K-means
Gang li

TF-IDF

review
(2010)

Movie

challenging. Our understanding and knowledge of the

78%

problem and its solution are still limited.

4

In future, more work is needed on further improving the

performance measures. Opinion Mining can be applied for
classificatio

many new applications. The techniques and algorithms used

Bayesian
Melville

91.21%

for opinion mining are very fast, and many of the study
n-gram
(2009)

blogs

Blog

remain unsolved. The main challenging aspects exist in the

use of opinion mining classifications, dealing with negation
5

expressions; produce a summary of opinions based on product

features, complexity of sentence/document, handling of
Document
frequency

hidden product features , etc. Many future research works

review
(2009)

Movie
hybrid
SVM,
Rudy

could be committed to these challenges.

89%
6

6. REFERENCES
[1] A. Berger, S. Della Pietra, and V.J. Della Pietra 1996. A
MI,JG, CHI
Songho tan

Chnsentico

90% SVM
classifier,

maximum entropy approach to natural language

Centroid
(2008)

processing , Computational Linguistics.

SVM

[2] B. Liu 2007. Web Data Mining, Exploring Hyperlinks,

rp
7

Contents and Usage data.

measureme
nt posts

[3] B. Liu 2011. Opinion Mining and Sentiment Analysis,

approach
Godbole

distance
Lexical

82.7%-

AAAI, San Francisco, USA.

(2007)

95.7%
Graph

Blog

[4] B.Liu 2010. Opinion Mining and Sentiment Analysis:

8

NLP Meets Social Sciences”, STSC, Hawaii.

Brill (2006)

[5] B. Liu 2008. Opinion Mining and Summarization, World

Wide Web Conference, Beijing, China.
n-gram
Konig,

review
hybrid

movie

91%

[6] B. Liu. 2010. Sentiment Analysis: A Multifaceted

9

Problem., Invited paper, IEEE Intelligent Systems.

Car review

[7] B. Liu. 2010 Sentiment Analysis and Subjectivity Second

Stemmed

Edition, The Handbook of Natural Language Processing.

Gamon
(2005)

Bayes
Naïve

terms

86%

[8] B. Pang, L. Lee, and S. Vaithyanathan, 2002. Thumbs

10

up? Sentiment classification using machine learning

techniques,” Proceedings of the Conference on Empirical
Hu and Liu

AmazonCn

DVD-73%

MP3-93%

Methods in Natural Language Processing (EMNLP), pp.

extraction
Opinion

Opinion

79–86.
(2005)

words

n. Net
word

[9] C. Cardie, J. Wiebe, T. Wilson, and D. Litman, 2003.

11

Combining low-level and summary representations of

opinions for multi- perspective question answering,
Lee (2004)

minimum
Based on
Pang and

review
Movie

86.4%
bayes,
Naïve

SVM

cuts
12

5
 International Journal of Computer Applications (0975 – 8887)
Volume 56– No.13, October 2012

Proceedings of the AAAI Spring Symposium on New [19] G. Jaganadh 2012. Opinion mining and Sentiment
Directions in Question Answering, pp. 20–27. analysis CSI Communication.
[10] ComScore/the Kelsey group 2007. Online consumer- [20] A.M..Popescu,, O. Etzioni, 2005. Extracting Product
generated reviews have significant impact On offline Features and Opinions from Reviews, In Proc. Conf.
purchase behavior, Press Release. Human Language Technology and Empirical Methods in
http://www.comscore.com/press/release.asp?press=1928 Natural Language Processing, Vancouver, British
Columbia, pp. 339–346.
[11] E. Riloff, and J. Wiebe, 2003. Learning Extraction
Patterns for Subjective Expressions, Proceedings of the [21] ZHU Jian, XU Chen, and WANG Han-shi, 2010.
Conference on Empirical Methods in Natural Language Sentiment classification using the theory of ANNs, The
Processing (EMNLP), Japan, Sapporo. Journal of China Universities of Posts and
Telecommunications, 17(Suppl.): pp. 58–62.
[12] T.Wilson,., J. Wiebe,., R. Hwa,. 2004. Just how mad are
you? Finding strong and weak opinion clauses. In: the [22] J. Martin 2005. Blogging for dollars. Fortune Small
Association for the Advancement of Artificial Business, 15(10), pp. 88–92.
Intelligence, pp. 761--769.
[23] J. Wiebe, E. Breck, C. Buckley, C. Cardie, P. Davis, B.
[13] H. Yu, and V. Hatzivassiloglou, 2003. Towards Fraser, D. Litman, D. Pierce, E. Riloff, T. Wilson, D.
Answering Opinion Questions: Separating Facts from Day, and M. Maybury, 2003. Recognizing and
Opinions and Identifying the Polarity of Opinion organizing opinions expressed in the world press in
Sentences, Proceedings of the Conference on Empirical Proceedings of the AAAI Spring Symposium on New
Methods in Natural Language Processing (EMNLP), Directions in Question Answering.
Japan, Sapporo.
[24] Christopher Scaffidi, Kevin Bierhoff, Eric Chang,
[14] W. Jin, H. Hay Ho, and R. Srihari, 2009. Opinion Miner: Mikhael Felker, Herman Ng and Chun Jin 2007. Red
A Novel Machine Learning System for Web Opinion Opal: product-feature scoring from reviews, Proceedings
Mining and Extraction. Proceeding of International of 8th ACM Conference on Electronic Commerce, pp.
conference on Knowledge Discovery and Data Mining 182-191, New York.
Paris, France.
[25] Yi and Niblack 2005. Sentiment Mining in Web
[15] L. Dey, S.K. Mirajul Haque, 2009. Studying the effects Fountain” ,Proceedings of 21st international Conference
of noisy text on text mining applications. Proceedings of on Data Engineering, pp. 1073-1083, Washington DC.
the Third Workshop on Analytics for Noisy Unstructured
Text Data, Barcelona, Spain. [26] M. Hu and B. Liu 2004. Mining and summarizing
customer reviews, Proceedings of the ACM SIGKDD
[16] B. Liu, and J. Cheng, 2005. Opinion observer: Analyzing Conference on Knowledge Discovery and Data Mining
and comparing opinions on the web, Proceedings of (KDD), pp. 168–177.
WWW.
[27] P.Turney 2002. Thumbs Up or Thumbs Down? Semantic
[17] G.Vinodhini and RM. Chandrasekaran 2012. Sentiment Orientation Applied to Unsupervised Classification of
analysis and Opinion Mining: A survey International Reviews. In: Proceeding of Association for
Journal of advanced Research in Computer Science and Computational Linguistics, pp. 417--424.
Software Engineering vol. 2 Issue 6.
[28] N.Mishra and C.K.Jha 2012. An insight into task of
[18] X. Ding, B. Liu, and P. S. Yu, 2008. A holistic lexicon- opinion mining Second International Joint Conference on
based approach to opinion mining, Proceedings of the Advances in Signal Processing and Information
Conference on Web Search and Web Data Mining Technology – SPIT.
(WSDM).