Dictionary Based Approach to Sentiment Analysis - A Review

International Journal of Advanced Engineering, Management and Science (IJAEMS) [Vol-2, Issue-5, May- 2016]
Infogain Publication (Infogainpublication.com) ISSN : 2454-1311
www.ijaems.com Page | 317
Dictionary Based Approach to Sentiment
Analysis - A Review
Tanvi Hardeniya 1
, Dilipkumar A. Borikar 2
1
M.Tech Student, CSE Department , Shri Ramdeobaba College of Engineering and Management Nagpur, India
2
Assistant Professor, CSE Department, Shri Ramdeobaba College of Engineering and Management Nagpur, India
Abstract— Due to the fast growth of World Wide Web the
online communication has increased. In recent times the
communication focus has shifted to social networking. In
order to enhance the text methods of communication such
as tweets, blogs and chats, it is necessary to examine the
emotion of user by studying the input text. Online reviews
are posted by customers for the products and services on
offer at a website portal. This has provided impetus to
substantial growth of online purchasing making opinion
analysis a vital factor for business development. To
analyze such text and reviews sentiment analysis is used.
Sentiment analysis is a sub domain of Natural Language
Processing which acquires writer’s feelings about several
products which are placed on the internet through
various comments or posts. It is used to find the opinion
or response of the user. Opinion may be positive, negative
or neutral. In this paper a review on sentiment analysis is
done and the challenges and issues involved in the
process are discussed. The approaches to sentiment
analysis using dictionaries such as SenticNet, SentiFul,
SentiWordNet, and WordNet are studied. Dictionary-
based approaches are efficient over a domain of study.
Although a generalized dictionary like WordNet may be
used, the accuracy of the classifier get affected due to
issues like negation, synonyms, sarcasm, etc.
Keywords— Sentiment Analysis, Natural Language
Processing, SenticNet, SentiFul, SentiWordNet,
WordNet.
I. INTRODUCTION
Human decision making or thinking is always affected by
others thinking, ideas and opinions. The growth of social
web gives a huge amount of user generated data such as
comments, opinions and reviews about products, services
and events. This data will be useful for consumers as well
as manufacturer. While buying any product online
consumers usually check comments or opinion of others
about the product. Manufacturer can understand the
response of that product and get insight into its products
strength and weaknesses based on the sentiment of the
customers. These opinions are helpful for both business
organizations and individuals but the huge amount of
such opinionated text data becomes burden to users. To
analyze and summarize the opinions expressed in these
enormous opinionated text data is a very interesting
domain for researchers. This new research domain is
typically called Sentiment Analysis or Opinion Mining.
Sentiment analysis is used to automatically mine the
opinions and emotions from text, speech, and database
sources with the help of Natural Language Processing
(NLP). Sentiment analysis does the classification of
opinions in the text into categories like “positive” or
“negative” or “neutral”. It’s often referred to as
subjectivity analysis, opinion mining and appraisal
extraction.
For buying any product customer wants to see the opinion
of other about the product. Customer reviews are very
important for business process since to make future
decision business organizations should know what
customers are saying about their product or service that an
organization is providing. It will provide important
functionality for voice of customer and brand reputation
management. Thus it is helpful in business process and
also for the customers
Major areas of research in Sentiment analysis are
Subjectivity Detection, Sentiment Prediction, Aspect
Based Sentiment Summarization, and Text
Summarization for Opinions, Contrastive Viewpoint
Summarization, Product Feature Extraction, and
Detecting Opinion Spam. Subjectivity Detection is a task
of finding whether text is opinionated or not. Sentiment
Prediction is about predicting the polarity of text whether
it is positive or negative. Aspect Based Sentiment
Summarization generates sentiment summary in the form
of star ratings or scores of features of the product. Text
Summarization generates a few sentences that summarize
the reviews of a product. Contrastive Viewpoint
Summarization puts an emphasis on contradicting
opinions. Product Feature Extraction is a work that extract
product feature from its review. Detecting Opinion Spam
is concern with identifying fake or bogus opinion from
reviews [5].

Sentiment classification is carried out at three levels
Document level, Sentence level and Aspect or feature
level. In Document level the task is to classify complete
document into positive or negative class. Sentence level
sentiment classification classifies sentence into positive,
negative, neutral class based on each sentence level. First
the polarity of each word of a sentence is calculated and
then the overall sentiment of the sentence is calculated.
Aspect or Feature level sentiment classification identifies
and extract product features from the source data and do
the classification [1].
Sentiment analysis is generally carried out by two
approaches: machine learning based and dictionary based.
Machine learning based approach applies classification
technique to classify text such as support vector machine
or neural network. Dictionary-based method uses
sentiment dictionary with opinion words and match them
with the data to determine polarity. They assign sentiment
scores to the opinion words describing the Positive,
Negative and Objective score of the words contained in
the dictionary.
Sentiment Analysis is generally done in two phase
subjectivity detection and polarity assignment. Firstly the
subject towards which the sentiment is directed is found
called subjectivity detection then, the polarity is assign
using dictionary such as SentiWordNet, WordNet,
SenticNet, SentiFul and others.
The issues in sentiment analysis are:
1. In different domain a positive or negative sentiment
word may have opposite orientations. For example,
“frozen” generally indicates negative sentiment for
software engineering but it can also imply positive
sentiment in air conditioning and refrigeration.
2. A sentence which contains sentiment words may not
imply any sentiment. This phenomenon occurs
frequently in several types of sentences. Question
(interrogative) sentences and conditional sentences
fall into this category. For example, “Can you tell me
which smart phone is good?”This sentence contain
the sentiment word “good” but does not state a
positive or negative opinion about the smart phone.
3. Many sentences which do not contain any sentiment
words can also express opinions. These sentences are
called objective sentences which are used to express
some factual information.
4. For example “This washer uses a lot of water”,
implies a negative sentiment about the washer as it
uses a lot of resource (water).
The natural language processing issues are:
1. Negation words are the words which reverse the
polarity of sentence. These are dealt with under
negation handling. For example, in the text “this
smart phone is not good”, the negation word “not”
reverses the polarity of sentence.
2. Word sense disambiguation is the lexicon ambiguity
that may be syntactic or semantic. It refers to the
words with more than one meaning that completely
different. For example, “I love this movie” and “This
is the love movie”. In this the word “love” has
different meanings.
This paper is organized as follows. Section 2 discusses
the current perspectives in sentiment analysis. Section 3
describes the main approaches to sentiment analysis. In
section 4 a detailed view of sentiment analysis using
dictionary-based approach has been deliberated. Section 5
covers the literature review on sentiment analysis. Section
6 concludes the discussion in earlier sections.
II. SENTIMENT ANALYSIS– CURRENT
PERSPECTIVE
Since form past few years sentiment analysis was done in
English language only. As many social networking sites
allow different languages for communication web content
are increased in a faster rate for other language also and
hence it is necessary to do sentiment analysis for other
languages. Recently Hindi opinion mining has been
carried out using two methods Machine learning method
uses Naïve-Bayes classifier. Part-Of-Speech tagging finds
adjective and consider it as opinion word and based on
their count document is classified. It uses TnT() POS
tagger and extract the adjective. The overall classification
accuracy is 87.1% [16]. Different language dictionaries
were also created for sentiment analysis. In recent times
sentiment analysis has been carried out in many
languages other than English. Sentiment analysis is not
only done on text data but also on visual images. For
extracting textual information embedded on images text
mining techniques incorporating on sentiment analysis
will be useful [17].Personality-based sentiment analysis is
performed as personality affect the ways people write and
talk and it is important for government and public
agencies to analyze the information propagating in social
networks. It provides higher accuracy value for both
positive and negative tweets than the baseline and
ensemble learning method [18].
An automated construction of the terminological
thesaurus for a specific domain is done. It uses
explanatory dictionary as the initial text corpus and a
controlled vocabulary related to the target lexicon to
begin extraction of the terms for the thesaurus. Sub-
division of the terms into semantic clusters is built on the
CLOPE clustering algorithm. It diminishes the cost of the
thesaurus creation by involving the expert only once

during the whole construction process, and only for
analysis of a little subset of the initial dictionary [19].
The values of precision and coverage depend on the
technique to build a lexicon. SentiWords a prior polarity
lexicon that is of approximately 155,000 words is
constructed that has both high precision and a high
coverage. It uses the experience of automatic derivation
of prior polarities from the SentiWordNet resource and a
collection of learning framework that take advantage of
manually built lexica [20].
III. SENTIMENT ANALYSIS– APPROACHES
Sentiment analysis approaches may be classified into –
machine learning approaches, lexicon-based approaches
and the hybrid approaches.
3.1 Machine Learning Approach
Machine learning approaches include Support Vector
Machine (SVM) and Naïve Bayesian classification. SVM
is a supervised learning method used to analyze the data
and recognize data patterns that can be used for
classification and regression analysis. Naïve Bayesian
Classification is based on Naïve-Bayes theorem and uses
the concepts of maximum likelihood and Bayesian
probability. The limitation of this method is that the
model needs to be trained with a large data volume before
testing. It is time consuming and low on accuracy when
training data is not sufficient.
3.2 Lexicon Based Approach
These approaches can be divided into two methods–
Dictionary based method, first finds the opinion word
from review text then finds their synonyms and antonyms
from dictionary. The dictionary used may be WordNet or
SentiWordNet or other. Corpus-based method helps to
find opinion word in a context specific orientation start
with a list of opinion word and then find other opinion
word in a huge corpus.
SentiWordNet 3.0 is most useful dictionary used. It is a
lexical resources publically available made up of
“synsets” each is associated with a positive, negative
numerical score range from 0 to 1. This score is
automatically allotted from the WordNet. It uses a semi-
supervised learning method and an iterative random walk
algorithm [6].
3.3 Hybrid Approach
It uses both the machine learning and the dictionary-based
approaches. It employs the lexicon-based approach for
sentiment scoring followed by training a classifier assign
polarity to the entities in the newly find reviews. Hybrid
approach is generally used since it achieves the best of
both worlds, high accuracy from a powerful supervised
learning algorithm and stability from lexicon based
approach [5].
IV. LEXICON–BASED APPROACHES IN
DETAIL
There are many lexicons available for sentiment analysis
such as SentiWordNet, WordNet, SentiTFIDF, SentiFul,
SenticNet, etc.
SentiTFIDF is based on proportional frequency count
distribution and proportional presence count distribution
across positively tagged document and negatively tagged
document and classify the term as positive or negative.
The term with equal proportion in positively tagged
document and negatively tagged document were
classified as a SentiStopWord and discarded. The process
is completed in three parts–calculating positivity of a
term, calculating negativity of the term, and classification
of the term based on its proportion of positivity and
negativity. SentiTFIDF has achieved accuracy of 92% [9].
SenticNet is a publicly available resource for opinion
mining construct based on Artificial Intelligence and
Semantic Web techniques. Dimensionality reduction is
used to deduce the polarity of common sense concepts
and hence provide sentiment analysis at a semantic level
rather than merely at syntactic level. It uses techniques
such as blending and spectral activation with emotion
categorization model and ontology for describing human
emotions. SenticNet is much more accurate than
SentiWordNet [10].
SentiFul database comprises of a reliable lexicon of
sentiment conveying terms, modifiers, functional word,
and modal operators. It gives a strong analysis on
orientation and strength of sentiment text. It differentiates
four types of affixes based on the role they play with
regards to sentiment features: propagating, reversing,
identifying and weakening. The sentiment conveying
words are found out through synonymy antonym and
hyponymy relations, derivation and compounding. It
helps to expand sentiment lexicon and improve coverage
of sentiment analysis [21].
Sentiment analysis using dictionary based approach is
described in Fig. 01 and may be performed as:

Fig. 1: A Framework for Sentiment Classification.
Initially input data is taken from the datasets then data
preprocessing is done. In this preprocessing extracts the
reviews and stop words are remove after that stemming is
done to convert the derived form of the word into root
form. Part-of-Speech tagging is done to assign the speech
to each word of the review so as to concentrate on the
adjectives, verb and adverbs. These words of review are
represented using n-grams. This representation is stored
in a database for sentiment polarity calculation. The
features from the database are retrieved and the sentiment
polarity is calculated using sentiment analysis technique
i.e. dictionary based technique. SentiWordNet dictionary
may be used for assigning the polarity to each word and
then the polarity of whole sentence is calculated. A pre-
defined threshold is used to compare the score value and
if it exceeds the threshold value it is classified as positive
otherwise negative. In situation when the difference is
zero, the text is classified to convey a neutral sentiment.
V. LITERATURE REVIEW
Emotion estimation is done using affective words and
sentence context analysis methods. It generates images
according to emotions in the text. For sentences that do
not contain the emotion words the emotion feature
estimation is done using sentence context analysis. This is
done by extracting the hidden knowledge of the sentence
by using Natural Language Processing Technique.
Emotion lexicon dictionary is used for extracting the
emotion feature that is made up of WordNet-Affect
database, WordNet 1.6 and SentiWordNet. The approach
is applied to two types of sentences – 1) Sentence with
direct emotional words for e.g. amazing, annoying.2)
Sentences with no emotion words but those contributing
for the emotion of writer [2].
SentiWordNet dictionary and smiley dictionary is utilized
to score the sentiment into positive, negative and
objective. Rule based and fuzzy logic approach is used to
handle negation words. Fuzzy Intensity Finder Algorithm
is used to find intensity of each word [3].
Sentiment analysis is carried out on Big Data as social
media available in internet is very vast. Hadoop is used to
perform sentiment analysis on large datasets and
performance of the system is measured. Lexicon based
technique is more appropriate then machine learning
algorithms is concluded for Big Data. Stemming is not
done in pre-processing as the dictionary is used which
contains all form of words. Negation and blind negation
words are found using dictionary and its polarity is
reversed. High speed for analysis of big data is present
but the accuracy is not too much [4].
The word which reverses the sentiment polarity of other
word is handled by a dependency tree based method for
sentiment classification using conditional random field
with hidden variables and analyzing interaction between
words [7].
A linguistic tree transform algorithm is used to eliminate
the word sense disambiguation and non-local dependency.
An objective sentence removal algorithm is used to
account the objective sentence. It performs better than n-
gram method [8].
Proportional frequency count distribution and
proportional presence count distribution are used for
sentiment analysis. SentiTFIDF used logarithmic
proportion of TFIDF of a term for positively tagged
documents and negatively tagged documents. If the
TFIDF of a term in positively tagged documents is larger
than TFIDF of same term in negatively tagged documents
the term is assigned positive polarity and vice-versa.
SentiTFIDF was more accurate than Delta TFIDF [9].
A common sense learning tool, SenticNet is a publicly
available semantic recourse for opinion mining built
using common sense reasoning technique together with
an emotion categorization model and an anthology for
describing human emotions. SenticNet is constructed
using ConceptNet and AffectiveSpace. SentiWordNet is
also incorporated. It is more superior to currently
available lexicon recourses [10].

Classification can be done on syntactic patterns and part-
of-speech tagging focusing on aspect level analysis using
SentiWordNet. This unsupervised method is a domain
independent method. SentiWordNet is applied using in
two phases. First phase SWN (AAC), considers
“Adverb+Adjective” combination and the other phase
SWN (AAAVC) considers “Adverb+Adjective”,
“Adverb+Verb” combinations. The method is applied for
finding out of sentiment polarity of all aspects in one
review [11].
A domain-independent lexicon based on Latent Dirichlet
Allocation for sentiment analysis is constructed. LDA is a
probabilistic model to construct a lexicon. The lexicon
constructed is highly related to the dataset. Precision of
this lexicon is more than the Liu’s lexicon, MPQA and
GI. This method is better than trivial methods in all
aspects as trivial approach builds the lexicon based on
calculating the words appearing number of times in
positive reviews and in negative reviews. [12].
Sentiments of product or services are different from social
issues sentiment. Verb plays an important role in analysis
of sentiments of social issue. The two main approaches is
used for sentiment analysis i.e. bag-of-words and feature-
based sentiment. A dictionary is constructed which
contain opinion verb, it has 440 opinion verbs. Some
Algorithms is given that extract opinions, construct
corresponding opinion structures, and at last calculate the
sentiment of opinion structures regarding the social issue
[13].
The different machine learning techniques and lexicon
based techniques, their limitations and the current
problem that researcher has studied in their work are
domain dependency; sentiment classification based on
insufficient labeled data; the lack of SA research in
languages other than English; and to deal with complex
sentences that requires more than sentiment words and
simple parsing [14].
A hybrid approach is mainly used for sentiment analysis.
It uses sentiment lexicon for polarity detection and the
results from the sentiment lexicon method are then use by
machine learning algorithms to train the data. It is
accomplishes through three phases, namely - crawler,
lexicon and machine learning module. It uses the AFFIN-
111 word list developed by FinnArup Nielsen. Feature
weighting method is applied since some of the features
had less impact on classification. Classification is done by
two algorithms Support Vector Machine (SVM) and K-
Nearest Neighbor (k-NN). It is observed than SVM
outperforms k-NN method [15].
VI. CONCLUSION
From the pioneering contributions to the domain of NLP,
IR and in specific to sentiment analysis it is observed that
sentiment analysis can play a very important role in the
development of successful business. Sentiment analysis
can be carried out in different languages and may extend
to other areas such as image processing, data aggregation,
etc. There are several dictionaries available for sentiment
analysis, of which SentiWordNet is used more often. This
paper reviews approaches, issues and challenges involved
in sentiment analysis and classification. Three types of
approaches are described with their relative merits
and limitations. It is found that sentiment analysis using
dictionary-based approach is swift compared to machine
learning-based approach since it require no prior training.
A framework of sentiment classification is explained
describing the main steps in sentiment analysis using
dictionary-based approach.
REFERENCES
[1] Varghese,R., Jayasree, M., “A Survey on Sentiment
Analysis and Opinion Mining,” IJRET: International
Journal of Research in Engineering and Technology,
Volume: 02 Issue: 11, October 2013.
[2] Patil, S. and T. Patil, “Automatic Generation of
Emotions for Social Networking Websites using Text
Mining,” 4th International Conference on
Computing, Communication And Networking
Technologies (ICCCNT) IEEE, 4-6, July 2013.
[3] Pimpalkar, A., Wandhe, T., Swati Rao M., Kene M.,
“Review of Online Product using Rule Based and
Fuzzy Logic with Smileys,” IJCAT – International
Journal of Computing and Technology, Volume 1,
Issue 1, February 2014.
[4] Kaushik, C.and Mishra A., “A Scalable, Lexicon
Based Technique for Sentiment Analysis,”
International Journal in Foundations of Computer
Science and Technology (IJFCST), Vol.4, No.5,
September 2014.
[5] Vohra, S. M. and Teraiya, J. B., “A Comparative
Study of Sentiment Analysis Techniques,” Journal of
Information, Knowledge and research in Computer
Engineering Volume – 02, Issue – 02, October 2013.
[6] Baccianella, S., Esuli, A. and F. Sebastiani,
“SENTIWORDNET 3.0: An Enhanced Lexical
Resource for Sentiment Analysis and Opinion
Mining,” The International Conference on Language
Resources and Evaluation (LREC). Vol. 10, 2010.
[7] Nakagawa, T., S. Kurohashi, “Dependency Tree-
based Sentiment Classiﬁcation using CRFs with
Hidden Variables,” Human Language Technologies:
The 2010 Annual Conference of the North American
Chapter of the Association for Computational
Linguistics. Association for Computational
Linguistics, 2010.

[8] B Eriksson, “Sentiment Classification of Movie
Reviews Using Linguistic Parsing. Natural Language
Processing,” CS, Volume – 838, 2006.
[9] Ghag, K., K. Shah, “SentiTFIDF – Sentiment
Classification using Relative Term Frequency
Inverse Document Frequency,” International Journal
of Advance Computer Science and Application,
Volume 5, No. 2, 2014.
[10]Cambria, E., Speer, R.Havasi, C. and A. Hussain,
“SenticNet: A Publicly Available Semantic Resource
for Opinion Mining,” In AAAI fall symposiums:
commonsense knowledge, vol. 10, p. 02, March
2010.
[11]Soni, V., Patel M. R., “Unsupervised Opinion Mining
from Text Reviews Using SentiWordNet,”
International Journal of Computer Trends and
Technology, volume 11 number 5, May 2014.
[12]Li, C. and R. Li, “Lexicon Construction: A Topic
Model Approach,” In International Conference
on Systems and Informatics (ICSAI), pp. 2299-2303.
IEEE, 2012.
[13]Karamibekr, M. and Ghorbani, A., “Sentiment
Analysis of Social Issues,” International Conference
on Social Informatics (Social Informatics), IEEE
2012, December 2012.
[14]Madhoushi, Z.,Hamdan, A. R. and Zainudin,
“Sentiment Analysis Techniques in Recent Works,”
Science and Information Conference (SAI), 2015.
IEEE.
[15]Mukwazvure, A., K.P. Supreethi, “A Hybrid
Approach to Sentiment Analysis of News Comments.
Reliability,” In 4th International Conference on
Infocom Technologies and Optimization (ICRITO)-
Trends and Future Directions, IEEE, 2015.
[16]Jha, V.,Manjunath, N., Shenoy, V.,Venugopal, K. R.
andPatnaik, L., “HOMS: Hindi Opinion Mining
System,” IEEE 2nd International Conference on
Recent Trends in Information Systems (ReTIS), M.
2015.
[17]Giannakopoulos, Theodoros, et al. “Visual Sentiment
Analysis For Brand Monitoring Enhancement,” 9th
International Symposium on Image and Signal
Processing and Analysis (ISPA), IEEE, 2015.
[18]Lin, Junjie, and M. Wenji, “Personality Based Public
Sentiment Classification In Micro-blog,” IEEE
International Conference on Intelligence and Security
Informatics (ISI), IEEE, 2015, pp. 151-153, M. 2015.
[19]Lagutina, Nadezhda, “An Approach to Automated
Thesaurus Construction Using Clusterization Based-
Dictionary Analysis,” 17TH Conference of Open
Innovations Association (FRUCT), IEEE, 2015.
[20]Gatti, L.,Guerini, M. and M. Turchi, “SentiWords:
Deriving a High Precision and High Coverage
Lexicon for Sentiment Analysis,” Affective
Computing, IEEE Transactions, 2015.
[21]Neviarouskaya, A., Prendinger, H. and Ishizuka, M.
“SentiFul: A Lexicon for Sentiment Analysis,” IEEE
transactions on affective computing, vol. 2, no. 1,
January-March 2011.

Dictionary Based Approach to Sentiment Analysis - A Review

More Related Content

What's hot (18)

Similar to Dictionary Based Approach to Sentiment Analysis - A Review (20)

Recently uploaded (20)

Dictionary Based Approach to Sentiment Analysis - A Review