Feature Based Semantic Polarity Analysis Through Ontology

IOSR Journal of Computer Engineering (IOSR-JCE)
e-ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 17, Issue 1, Ver. IV (Jan – Feb. 2015), PP 13-17
www.iosrjournals.org
DOI: 10.9790/0661-17141317 www.iosrjournals.org 13 | Page
Feature Based Semantic Polarity Analysis Through Ontology
Dr.M.S.Anbarasi1
, Sarathkumar.B2
, Bomanna jaswanth2
,
Shahul Nikamuthul asiqine.H2
1( Department of Information Technology , Pondicherry Engineering College, Pondicherry ,India. )
2( Department of Information Technology , Pondicherry Engineering College, Pondicherry ,India.)
Abstract: Opinion mining, a trending research area where customers feels that opinions of others are always
important for making decisions while purchasing the products. Here the problem is to collect those opinions
and preprocess them. Thus, the customer reviews have been collected and after that Ontology is constructed to
structure the information available in unstructured text reviews and to exploit semantic relation between them.
Then we extract the features of the product and their opinions further polarities (positive, negative or neutral) of
different features of certain digital products are identified. Opinions of different features of productivity
customer are classified and summarized by an enhanced opinion mining technique. The performance of the
system is evaluated by metrics such as precisions, recall and F-measure. This information provided to users will
be more helpful to make decisions before buying a product.
Keywords: ontology,opinion mining,feature extraction,pos tagging,polarity identification
I. Introduction
Opinion mining is a highly trending research field that comprises natural language processing,
computational linguistics and text analysis techniques with the aim of extracting various kinds of added-value
and informational elements from users‟ opinions. The task of organizing user feedback and retrieving
information is known as opinion mining. Customers‟ satisfaction is the key to the manufactures for the success
of their business, the opinions (or) feedbacks which are given by the customers are necessary to know about
their products in the market.
Approaches of opinion mining where user reviews has been collected and classified into two
approaches: text classification and information extraction approaches. In the former, researchers have been
exploring techniques for polarity identification for entire document as either positive vs. negative [Bing Liu
2008; Pang and Lee, 2004; L.Zhuang 2006, etc.]. The latter, on the other hand, describes about the products or
reviews at feature level, which helps both customer and developer to know in detail about their strength and
weakness.
Time is running from pin to pin, people preferences to the online shopping is comparatively more than
generic method of in store shopping. Supporting to this many review sites and online websites is helping to the
customers with consisting of the feedbacks and reviews.
Many researchers are being focusing on polarity identification of user reviews from years even though
the accuracy is not achieved. Every document consists of both positive and negative feedbacks ,according to
many opinion mining approaches the document which consisting most positive preferable sentences is considered
as positive feedback ,in the same way for negative feedback. The former approach (Zhou, L., & Chaovalit, 2008;
Neviarouskaya, Prendinger, & Ishizuka, 2009; Taboada, Brooke, Tofiloski, Voll, & Stede, 2011) is based on
opinion words and extracting information about the positive ,negative and neutral documents.
In fact, opinions classified at the document or sentence level determines the rating of the product or
whether the product is good, bad or neutral. A product which is identified by these mining techniques as good
does not mean that the user has positive opinions on all aspects or features. Likewise, a bad product does not
mean that the user dislikes everything about the object. In a document (e.g., a product review), the user
typically writes about both the positive and negative aspects of the object, although the general sentiment toward
that object may be positive or negative (Ahmad & Doja, 2012). To obtain such detailed aspects, it is necessary
to perform feature-based opinion mining in an attempt to identify the features in the opinion and classify the
sentiments of the opinion for each of these features (Feldman, 2013).
An ontology can be defined as an „„explicit, machine-readable specification of a shared
conceptualization‟‟ (Studer, Benjamins, & Fensel, 1998). Many authors utilizing ontologies for defining a
domain so that it serves many purpose. As well as the relations among these terms and are now applied in
various fields, like agent and knowledge management systems and e-commerce platforms (Gómez-Pérez &
Corcho, 2002).

. In this paper we propose an innovative opinion mining methodology that takes advantage of new
Semantic Web-guided solutions and ontology to enhance the results obtained with traditional natural language
processing techniques and sentiment analysis processes. In this paper first we will collect the customer review
data, those data will be processed. Further we will extract the features through pos tagging and frequency
calculation. Finally the main goal of all these works is to taking into account about each aspect of the products,
classifying each feature are important and giving polarities and semantic scores to it .
II. Headings
1. Data Collection
The basic step we should perform is collection of data.To determine the polarity of the sentences, based
on aspects, large numbers of reviews are collected from the Web. There are lots of websites on the Internet
where the large numbers of customer reviews are available. We used One of the popular website in social
network is Amazon (www.amazon.com) to collect the reviews from web.
2. Data preprocessing
Data processing should be done to those reviews which are having been collected from the website. But
in those reviews some unwanted information is available. We should remove those information like URL links
(e.g. http://semantic.com), Twitter user names (e.g.@lifestyle – with symbol @ indicating a user name),Twitter
special words (such as “RT”6) and emoticons. But those information with “RT” should be collected, RT means
retweet which is an useful information, an aspect that more retweets means more people are discussing about
that aspect. User reviews can be a mix of both subjective and objective sentence. Subjective sentences are
mostly based on facts so it should not be considered for the further process, and only objective sentences are to
taken for the remaining process. Then pos tagging module is done for the extraction of the features which is the
next step of the process. A tool of the name Stanford POS tagger tool
(http://nlp.stanford.edu/software/tagger.shtml) is utilized for tagging purpose. It is a piece of software which is
used to read those human reviews and assigns parts of speech for each word of a sentence. So, that feature
extraction is possible and then Term frequency, dispersion, deviation is to be calculated. These terms are useful
in determining the polarities of the features.
III. Frequency calculation
The TF/IDF (TF-Term Frequency, IDF-Inverse Document Frequency) algorithm is the best algorithm
which is well known for weighting scheme for terms in the information retrieval. The TF/IDF score (or, weight)
which depends upon the occurrences and it is also considers the length of the document and the query terms, it is
nothing but a term increases with the number of occurrences in a document (TF component). The score will also
increase with the rarity of the term across the entire collection (IDF component) and
3.1Document Frequency (DF)
The ratio of the document that contains the term t to the total number of documents that contains term t
3.2Inverse Document Frequency (IDF)
The IDF component acts to discriminate between informative and non-informative query terms. Those
terms that have a high IDF are considered more informative, because they rarely occur in the collection. On the
other hand, terms that have a low IDF are considered uninformative, since they occur in many documents. As
the number of documents in a collection increases, IDF becomes increasingly important in order to discriminate
between those documents that contain non-informative query terms and those that contain high informative
query terms.
This value is obtained by the taking the total number of documents in the collection and then differentiating
then dividing it with the former gives the value of idf
idf = log10( | D | / dft )
3.3 Term Frequency (TF)
The TF component, which is often normalized in some way with respect to the document length, is
used to discriminate between documents that contain a query term several times and those that contain the term
many times. This makes the assumption that documents that contain more mentions of a given query term and
are more "about" the given term and therefore are more likely to be relevant to the query. Note that this can be a
poor assumption particularly when collection size increases and it becomes nosier .Repeated words might be
because of irritation towards the product The TF component becomes more important when documents grow
longer that the length of the document extended.

This measurement is a simple one:
1 + log10(tft, d)
IV. Ontology Creation
Creating Ontology is long and time consuming process which is also requires more human effort for
this creation. To reduce the time and effort for this length process of ontology, the ontology Learning has been
developed which is useful to know the concepts and relations that exists in particular domain. This will be
achieved with the use of certain tools like natural language (NLP) process tools which includes a morphologic
analyzer, a part-of-speech (POS) tagger, and a chunk parser, unstructured customer statements on a specific
product can be translated, analyzed and modeled automatically.
Ontology applications include natural language generation, intelligent information integration,
semantic-based access to the Internet and extracting information from texts. However, the most important
contribution of ontologies is the key role they play in the development of the Semantic Web. The Semantic Web
is an extension of the current Web, where information is given a well-defined meaning, encouraging cooperation
among human users and computers (Berners-Lee, Hendler, & Lassila, 2001). Ontologies serve as the primary
means of knowledge representation in the Semantic Web. Although various ontology languages have emerged,
the currently dominant standards are RDF/S (Resource Description Framework Schema) and OWL (Web
Ontology Language). Additional points of motivation for preferring the use of ontologies in an application
include: (a) analyzing domain knowledge and separating the latter from operational knowledge, (b) Enabling the
reuse of domain knowledge, (c) Making domain assumptions explicit, and (d) Sharing a common understanding
of the information structure among people and/or software agents.
Ontology construction makes us to know the various levels of abstraction through the customer
statements about the characteristics of the product. Construction of ontology can be done both on manually and
on machine learning. It can be done with the help tools which include editors, checkers to extract relation among
the entities of the product.
The first step in ontology creation is classifies the attributes it is nothing but it extracts the features of
the product. Then it will proceeds with the sub groups (or) other attributes to existing attributes as features, it
forms an ontology structure in the hierarchical form with the different aspects of the products. There is also
chance of existing similar text repeating in a review those are not being considered in the classification. In the
available text document it also makes text into three categories of structured text, unstructured text and semi
structured text. Each word in the review is verified and classified. After ontology creation then it provides
semantic score to each aspect with the various enhanced opinion mining techniques. It also makes the domain to
reuse .
V. Opinion mining
After analyzing the corpus by means of NLP techniques, extracting the relevant features and
identifying the features polarity, the framework proposed here provides an innovative opinion mining
mechanism. The opinion mining module described in this section is based on vector analysis and enables an
effective feature sentiment classification. In the opinion mining process it generates the opinions about the
products based on the review analysis and polarities.
After extracting the features of the product through the ontology then we should semantic score and
polarities i.e. either it can be positive (or) negative (or) neutral. Those semantic scores are assigned scores with
the help of n-grams technique.
There are various tools available for providing polarities to the features of products, among them one
of the best tools Sentiword.Net is used in this paper. The tool provides polarities to the words of the reviews.
Positive, Negative and neutral score are being calculated with the average words that are present in the review,
if more positive words it is considered as positive score, consists of negative words is considered as Negative
score and neutral score is defined with their sum less than one.
N Gram‟s technique of opinion mining is utilized for to identify that how many words nearly and
prefers to the features of the products. It consists of N_Gram After, N_Gram Before, N_Gram Around which
gives the detail which specifies the features of the products. So, that accurate semantic score as being calculated
semantic score and polarities. The words which indicates positive is considered as supporting to feature and
negative words which are blaming considers the negative words.
We will provide the score on a scale -5 to 5. If the calculated score from range below zero it indicates
Negative score and polarity will be negative. If the score is zero then it is neutral polarity, and the score is more
than zero it is Positive score. The customers are being satisfied with the semantic score which is provided with
this paper.

Since the polarity is calculated for each features separately the techniques of N grams is applied to the
collected words of each features in the structure that is created by the ontology .To provide a sentiment score for
each words ,we use sentiwordnet . The feature polarity is calculated using SentiWordNet 3.0 (Baccianella &
Esuli, 2010) (SWN). SENTIWORDNET is the result of the automatic annotation of all the synsets of
WORDNET according to the notions of “positivity”, “negativity”, and “neutrality”. Each synset s is associated
to three numerical scores Pos(s), Neg(s), and Obj(s) which indicate how positive, negative, and “objective” (i.e.,
neutral) the terms contained in the synset .Different senses of the same term may thus have different opinion-
related properties.
INTENDATIONS and EQUATIONS
Algorithm
Input: User reviews
Output : Opinion Results and ontology
Process :
Tagging process  POS TAGGING
Extraction of Features f1,f2,f3……fn;
Word count=Sum (Number of words in each document);
For each word w
Term Frequency=Number of occurrences of each word w/Total number of words in the
document.
End loop
For each word w
Opinion Strength Calculation = Sign(SET(p))(CS(p)/Set(p))
Inverse Document Frequency=ln(R/RCA)*gamma
Gamma=1/ln(R)
Sentence=OpinionStrength*InverseDocumentFrequency*
Degree
End loop
For each word w
Opinion Score=Averagepolarity*populationweight*weight
End loop
Output ontology file.
VI. Figures and Tables
Reference
[1] EfstratiosKontopoulos,ChristosBerberidis,Theologos Dergiades, Nick Bassiliades, “ Ontology-based sentiment analysis of twitter
posts,“Expert Systems with Applications 40, Elsevier (2013) 4065–4074.
[2] Abd. Samad Hasan Basaria,Burairah Hussina,I. Gede Pramudya Anantaa, Junta Zeniarjab, “ Opinion Mining of Movie Review
using Hybrid Method of Support Vector Machine and Particle Swarm Optimization”, Elsevier, Procedia Engineering 53 ( 2013 )
453 – 462.
[3] Li Chen, Luole Qi, Feng Wang, “ Comparison of feature-level learning methods for mining online consumer reviews”, Elsevier,
Expert Systems with Applications 39 (2012) 9588–9601.

[4] Baldoni, M., Baroglio, C., Patti, V., & Rena, P. (2012). From tags to emotions Ontology-driven sentiment analysis in the social
semantic web. Intelligenza Artificiale, 6(1), 41–54.
[5] De Kok, D., & Brouwer, H. (2012). Natural language processing for the working programmer. E-book available under the creative
commons attribution 3.0 License (CC-BY). 2011,access: November 2012.
[6] Moreo, A., Romero, M., Castro, J. L., & Zurita, J. M. (2012). Lexicon-based commentsoriented news sentiment analyzer system.
Expert Systems with Applications, 39(10), 9166–9180.
[7] Zhang, L., Liu, B., Lim, S., & O‟Brien-Strain, E. (2010). Extracting and ranking product features in opinion documents. In
Proceedings of the 23rd international conference on computational linguistics: Posters (pp. 1462–1470). Association for
Computational Linguistics .
[8] Ning, L., Guanyu, L., & Li, S. (2010). Using formal concept analysis for maritime ontology building. In Proceedings of the 2010
international forum on information technology and applications (IFITA „10) (pp. 159–162). New York, USA: Springer.Vol.2

Feature Based Semantic Polarity Analysis Through Ontology

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Feature Based Semantic Polarity Analysis Through Ontology (20)

More from IOSR Journals (20)

Recently uploaded (20)

Feature Based Semantic Polarity Analysis Through Ontology