CiiT International Journal of Data Mining and Knowledge Engineering, Vol 8, No 07, July 2016 233
0974-9683/NCRTIT16-042/05/$20/$100 © 2016 CiiT Published by the Coimbatore Institute of Information Technology
Abstract---The usage of the internet has improved rapidly in the
recent years. Opinion mining intends to examine opinion of a
product, services and its attributes. The opinions extracted are in
terms of sentiments, subjectivity of text and attitude. Opinion mining
is research field of mining textual data from the web. There is a vast
increase of textual data in the web. The customers are given an
opportunity to write their own opinion on the products they purchase
online and use. To categorize these opinions and to find the polarity,
opinion mining is used. There are various models, techniques and
algorithms proposed for mining these opinions. In this paper, we have
examined the latest opinion mining algorithms. These algorithms are
evaluated based on their performance. A detailed study has been done
which gives an idea of recent algorithms used.
Keywords---Opinion Mining- Techniques- Algorithms
I. INTRODUCTION
PINION mining gathers and classifies opinion regarding
the product. Opinion mining is also known as
Sentimental Analysis. Opinion mining is useful in number of
ways. It assists vendors to assess accomplishment of the newly
launched product in the market. It decides which version of
the product is well-liked by the customers. It extracts the
opinions of users which they provide on a particular product
[1-4].
The evolution of opinion mining has been put into five
phases. Those five phases were described using the fish-bone
diagram.
Manuscript received on May 19, 2016, review completed on May 19, 2016
and revised on May 28, 2016.
U. Prabu is with the Department of CSE, Shri Krishnaa College of
Engineering and Technology, Pondicherry, India. E-Mail:
uprabu28@gmail.com
P. Balasubramanian is with the Department of CSE, Shri Krishnaa College
of Engineering and Technology, Pondicherry, India. E-Mail:
pbalasubramanian02@gmail.com
M. Sithanandam is with the Department of CSE, Shri Krishnaa College of
Engineering and Technology, Pondicherry, India. E-Mail:
sithu.cse.235@gmail.com
A. Anidha is with the Department of CSE, Shri Krishnaa College of
Engineering and Technology, Pondicherry, India.
Digital Object Identifier: DMKE072016005.
Fig. 1. Five Phases of Opinion Mining Evolution
Sentimental analysis is done in three different levels
namely: Document level, Sentence level and Phrase level. In
document level analysis, objective and subjective sentence
classification is much important. Extraneous sentences are
removed during processing. The unsupervised and supervised
learning methods are used in document level analysis. In
sentence level analysis, each sentence's polarity is computed.
The positive and negative classes are the two classes of the
polarity classification. Phrase level analysis is the identity to
opinion mining approaches.
The tools used in opinion mining for tracking the polarity
are Review Seer tool, Web Fountain, Red opal and Opinion
observer.
The opinion mining tasks are Subjectivity and Polarity
classification, Opinion source identification, Opinion target
identification and Opinion summarization.
The opinion mining techniques are supervised machine
learning, unsupervised machine learning and Case based
reasoning.
A. Motivation
There are quite a few algorithm existing for opinion mining.
Each of these algorithms are enclosed by considering certain
inputs from various resources. Some algorithm concentrate on
the real time opinion whereas some concentrate on the datasets
available. So by extracting these features we can obtain the
knowledge of opinion mining algorithms.
B. Contribution
A comprehensive study on opinion mining algorithms is
presented in this paper. Opinion are having a major part in
today's electronic business world. Without the opinion and
reviews the market value of a product can't be determined. The
opinion mining algorithms are discussed in depth. To the best
of our knowledge, we admit that this a first comprehensive
study on recent opinion mining algorithms and their
performance evaluation.
A Comprehensive Study and Performance Evaluation
of Opinion Mining Algorithms
U. Prabu, P. Balasubramanian, M. Sithanandam and A. Anidha
O