Ontology-Based Opinion Mining

Introduction

Opinion Mining, also known as sentiment analysis, is a process which aims to determine the polarity of a textual corpus i.e document, paragraph or sentence etc., ending towards positive, negative or neutral sense. Economic and Financial modeling could be one of the most promising application of Opinion Mining/Sentiment Analysis. Machine Learning based sentiment classifiers also consist of an opinion mining tool. But the main problem with these classifiers is that they classify opinions of users into classes (+ve, -ve or neutral) which assigns a corresponding score to each of the class as a whole rather than considering that many aspects of the same notion is also there. Machine learning approach always gives a single quantitative (sentiment score) or qualitative result (+ve or –ve). So, there is a requirement of ontology based techniques which is a fine grained version of Opinion Mining/Sentiment Analysis.

Ontology is an explicit detailed study of concepts. The word ‘Ontology’ is borrowed from philosophy. Ontology can be defined as an “explicit, machine-readable specification of a shared conceptualization’. It is generally considered as a formal specification of conceptualization which consists of concepts and their relations. It can also be defined as systematic maintenance of information about the things which already exist.

In Computer Science, “Ontology Based Opinion Mining” is a formal representation of knowledge with the help of a fixed set of believed concepts and the relationship between those concepts. It could have the reasons about the domain’s properties and with the help of that property a domain can be described easily. So, it is a formal explicit description of the basic/core concepts of domain. It also provides a shared word stock/lexicon.

It is an identification and extraction of subjective material from different types of source material or from the unstructured data with the help of natural language processing, computational linguistics and text analytics.

Ontologies are used in many areas like Artificial Intelligence, Library Science and Information Science. Mainly there are 4 categories of ontologies:

• 1.

  Static: This type of ontology tells us about the existing things, their attributes and relationship between them.

• 2.

  Dynamic: It describe transition, process and morphisms of various types of things.

• 3.

  Intentional: It deals with want, prove or disprove and argue about the things as per user’s own perception.

• 4.

  Social: It relates with surrounding environment and its settings.

Ontologies can be used in many ways. Within Computer Science, it is a model for describing world which has a particular set of types, their relationships and properties. It can also be assumed that there is a very thin line of difference between real world and the model features used for ontologies. Text based online reviews by customers for a product is fair and unbiased parameter to quality measure of the product. But the problem is that all these reviews cannot be read by all users. So we have to apply some techniques of opinion mining which can construct a feature specific graphical representation and give us the real picture of relative quality of the product.

The main aim of this is to find out mind set or thought process of a speaker or a writer related to the relevant topic or polarity of the document in terms of context. Mind set could be his/her own judgment or affective state or predefined emotional communication.

As per the form of opinions, they can be classified into three goups:

• •

  Form 1: Advantage, Disadvantages and Summary of Opinions.

• •

  Form 2: Advantage, Disadvantages of Opinions.

• •

  Form 3: Free Form, (No Rules and Regulations for specification of forms).

Literature Review

This section is related to the work carried out in the past on Sentiment Analysis and Opinion Mining. In this chapter, various techniques of Sentiment Analysis and opinion mining will be discussed in detail. Opinions can be extracted with the help of extraction of structured information which could be entities, their attributes and relation among them.