A REVIEW PAPER ON OPINION MINING AND SEMANTIC ANALYSIS

Prof. R. B. Ghate¹, Prof. A. N. Thakare²

¹Assistant Professor, Computer Engineering Department, BDCE Sevagram

² HOD of Computer Engineering Department, BDCE Sevagram

Abstract:

Opinion mining also referred as Sentiment Analysis, it can be determined as the task of discovering, deriving and classifying Opinions on anything. Opinion mining is essential for e commerce websites, blogs, customer reviews on products, social media data, online portals, web forums and policies, Peoples reviews on these are increasing day by day. It is very problematic to keep a record of the opinions of the peoples. Sentiment analysis finds the abstract information from the source information by using NLP. This review paper focuses on published literature on approaches, applications, challenges, algorithms and tools of sentiment analysis and opinion mining.

Keywords: Opinion mining, Sentiment analysis, NLP, Text Mining.

Introduction:

The growth of computerized systems and digital information in every field of day to day life is emerging speedily. As a result immense data are generated in field of technology, medical, marketing, finance, demographic, etc. Opinion mining also called as sentiment analysis is a process of identifying people’s important opinion about a particular topic like news, movies, tourist places, foods galleries, online web contents or product and problem. At the point when any choices are to be made in regard to the buying of new item, hardware, software or electronic items the individuals are a lot of keen on getting the surveys of the different sites, websites or conversation gatherings. In such case opinion mining or sentiment analysis is utilized broadly which manages the state of mind of the individuals with respect to a specific item or product. Opinion mining is a topic deals with the area of Text mining, Natural Language Processing, and Web mining. The aim of Opinion Mining is to make the computerized system able to notice and express emotions.  An Idea, view, or belief based on feelings or emotions rather than reason is called sentiment [1][2].

Opinion mining defined as:  Opinion mining (sentiment mining, opinion/sentiment extraction) is the area of research that attempts to make automatic systems to determine human opinion from texts written in natural language.  It is the process of analyzing the text about a topic written in a natural language. Opinion mining classifies them as Negative, Positive or Neutral, based on the peoples sentiments, emotions and opinions expressed in it. These reviews are used by the individuals and the organization for decision making.  The usefulness of opinion mining is increasing day by day.

 Levels of Opinion Mining:

• Document Level
• Sentence Level
• Aspect Level
• Document Level:

In the document level task complete document of opinion mining is categorized into two sentiments, positive sentiment and negative sentiments [4].

• Sentence Level:
  In the sentence level task each sentence takes into consideration, whether each sentence represents a positive, negative, or neutral opinion. Neutral opinion represents no opinion [4].
• Aspect Level:
  In the aspect level task neither document, nor sentence takes into consideration, it looks only the opinion. Opinion considers positive, negative, neutral and target. Aspect level sometime called as feature-based opinion mining [4].
• Sentiment Analysis:
  Sentiment analysis is a part of Natural Language Processing where emotions or opinions of individuals are undermined.  The texts from reviews or opinions are processed to get an accurate description of how the individual feels regarding the subject.

Phases:

• Preprocessing phase:
  The information entered by the client is first cleaned to minimize the noise so that key words could be analyzed.
• Feature Extraction Phase:
  In this stage the key words are given a token and are put under analysis process.
• Classification Phase:
  In this stage all the data is classified. Based on different algorithms. 

SA is important to examine the every opinion and determine the following:

• Subjectivity:
  The content is analyzed to decide whether it is subjective in nature, implies the analyze content contain any opinionated contents [5].
• Polarity:
  The Polarity is classified into two sorts i.e. positive or negative opinion [5].
• Polarity Strength:
  The polarity can be categorized into diverse levels. Polarity strength can be can be seen as strongly positive, slightly positive, weekly positive, weekly negative, moderately negative, negative, and strongly negative [5].

Literature Survey:

Author Bilal Saberi, Saidah Saad done the Survey on Semantic analysis and Opinion mining [9] in this paper author states the various approaches available for Semantic analysis. An approach consists of Machine Learning, Lexicon-Based and Combination Method.  

Polarity of sentiment is presented by author Jyoti and Seema Rao [10] they also compare the two available tools Weka and Mahout for sentiment analysis.

[2] In this paper author explained certain algorithms that are commonly used in sentimental analysis and opinion mining such as Support vector machine-nearest neighbour, K-means clustering.

B.Sampath Kumar [1] in this paper analyzes the various techniques for semantic analysis and opinion mining as well as analyzes the various tools for the semantic analysis and opinion mining

Approaches [9]:

• Machine Learning Approach:

There are multiple techniques in this group   employed in an effort to gain more salient features. Give accurate information regarding the polarity of Sentiments.

Various ML Approaches:

• Support Vector Machines (SVM)
• Naive Bayes (NB)
• Maximum Entropy (ME)
• K-Nearest Neighbour
• Lexicon-Based Approach:

The second group uses a language-inclined approach named the approach focused on lexicon. The analysis is begins with words or sentences representing characteristics of semantic polarity.

• K-means clustering method is used.
• Combination Method Approach:

Here is also another community, which is used to combine machine learning with group lexicon.  That’s it.   Combination system community called or semi-supervisor.

• Improved Naïve Bayes and SVM algorithms are used.

Techniques [10]:

Categorization of sentiment analysis is done by these techniques.

1. Machine learning technique

• Supervised learning
  SVM (Support Vector Machine), NB (Naive Bayes), ME (Maximum Entropy) these methods are used for classifying the users review in supervised learning technique.
• Unsupervised learning
  K-Means Clustering method is used is by researcher to classify the opinions about the products. TF-IDF weighting method also used.

2. Natural language processing technique

To extract the keywords, entities and sentiments Alchemy API is used. Star rating system solves the fake opinion problem. N-Gram and part of speech extraction is used for pattern identification. Then K-L divergence algorithm is used for removal unwanted comments. Fuzzy logic based sentiment analysis also used in NLP.

Algorithms [2]:

A. Classification Algorithm

1. K-Nearest Neighbor Algorithm:
   This algorithm is used for classification and regression.
2. Support Vector Machines Algorithm:
   SVM Algorithm used for classification, regression and pattern recognition.

 B. Clustering Algorithm

1. K-means clustering algorithm:
   This is partitioning algorithm, used for contraction and evaluation of different patterns.
2. Self organized Map (SOM) algorithm:
   This is an artificial neural network (ANN) type of a network. SOM algorithm used to detect the features that are inherent to the Problem.

Challenges [5]:

Challenges in Opinion Mining and Sentiment Analysis:

1. Authority
2. Non-Expert Opinion
3. Spam Opinion Detection
4. Biased or Spam Opinion
5. Opinion Credibility
6. Domain Dependency of Sentiment Analysis
7. Divergent from the Topic (Opinion Relevancy)

Tools [1][5]:

Tools used for Sentiment analysis and Opinion mining

1. Review Seer tool
2. Web Fountain
3. Red Opal
4. Opinion observer
5. Opinion Finder
6. Natural Language Tool Kit (NLTK)
7. Stanford Parser
8. Ling Pipe
9. Stanford CoreNLP
10. SVM multi-class

Applications [4]:

The applications of Opinion mining and sentiment analysis are as follows:

1. To buy a Product or Service
2. Recommendation Systems
3. Quality Improvement
4. Detection of “flames”
5. Marketing research

Conclusion:

Opinion mining and sentiment analysis is a field of extraction of knowledge from various reviews given by people or user on product or anything.  This paper is giving the overview regarding the opinion mining and sentiment analysis and approaches, algorithms, techniques, tools and challenges.

References:

1. B. Sampath Kumar, “An Analysis on Opinion Mining: Techniques and Tools”, Indian       Journal of Research, Volume: 5 | Issue: 8 | August 2016
2. G. Sneka, CT. Vidhya, “Algorithms for Opinion Mining and Sentiment Analysis: An Overview” IJARCSSE, Volume 6, Issue 2, February 2016
3. Ion Smeureanu, Cristian Bucur, “Applying Supervised Opinion Mining Techniques on Online User Reviews”, Informatica Economică vol. 16, no.  2/2012
4. Dhanashree Kulkarni, Prof S.F.Rodd, “A Survey on Opinion Mining problem and levels of Analysis”,IJIRSET, Vol. 4, Issue 12,  December 2015
5. Sarwar Shah Khan, Muzammil Khan, Qiong Ran and Rashid Naseem, “Challenges in Opinion Mining, Comprehensive Review”, Science and technology journal, Vol. 33 No. 11. PP. 123-135, Nov 2018
6. A’sim Seedahmed Ali, “Opinion Mining Techniques”, IJISET – International Journal of Innovative Science, Engineering & Technology, Vol. 2 Issue 6, June 2015.
7. Santhosh Kumar K L, Jayanti Desai, Jharna Majumdar, “Opinion Mining and Sentiment Analysis on Online Customer Review”, IEEE International Conference on Computational Intelligence and Computing Research, December 2016
8. Dr. B. Radha, V. Meera, “A Survey on Research Issues in Opinion Mining”, IJIRCCE, Vol. 5, Issue 2, February 2017
9. Bilal Saberi, Saidah Saad, “Sentiment Analysis or Opinion Mining: A Review”, International Journal on Advanced Science Engineering Information Technology, Vol.7 (2017) No. 5   
10. Jyoti, Seema Rao, “A Survey on Sentiment Analysis and Opinion Mining”, ACM. ISBN 978-1-4503-4213-1/16/08, 2016
11. Kumar Ravi, Vadlamani Ravi1, “A survey on opinion mining and sentiment analysis: tasks, approaches and applications”, Elsevier, Knowledge-Based Systems Vol 89, November 2015, Pages 14-46