Anda di halaman 1dari 7

IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

Comparitive Analysis of Various Type of Quality Measuring


Techinque of Ontologies

Sameer Dass#1
#Computer Science Department, Ambedkar Institute of Advanced Communication Technologies and Research, NewDelhi.
1dass.sameer2@gmail.com

Abstract

Ontologies are utilized in numerous fields which includes expertise management, facts extraction and semantic net. Moreover this
paper also focus on the various ontologies selection technique for ontology based on the context based or theme. The reuse of
ontologies without anomalies is a vital factor in the business place so as to produce successful tasks; therefore the selection of an
assessment technique is obligatory. This paper provides the contrast of numerous evaluation strategies based totally on specific
parameters with the intention to assist the users. There are fixed software criterion to judge whether it is healthy or not ontologies to
solve any particular problem.

Keywords Ontology, Schemas, Owl, Meta data, Cohesion, Richness

I. INTRODUCTION
This document is a template. The inventor of the World Wide Web, Tim Berners lee and his world wide web Consortium team are
working to develop, extend and standardize the webs mark-up language and tools. The objective of the semantic web structure is to
offer an understanding representation of related data for you to permit the system processing the on an international scale .The
semantic web may be expected to capitalize the machine understandable language. For better performance we need to create better
ontology and there evaluation and measuring of ontology both during in creation and application purpose are very important .The
main technology of semantic web such as XML, RDF and ontologies. Ontologies is an information about the nature of a entity .Web
researchers have adopted the term own purpose and they defined as a document file which contain formal relations among terms and
entity. They use ontology languages such as RDF Schema [4] and OWL [5]. These provide the common model data representation for
integration and aggregation.

Current ontology researches and practices have been used in artificial intelligence representation, they use the technology like OIL,
DAML, OWL, KIF, KL-ONE etc. In recent year there are numerous technique are available for ontology evaluation and
suggestion .Some author suggest that metrics evaluation are easy and very informative. Mostly method work on the structural
annotation of ontologies .In the paper we compare various method of ontologies evaluation that can be help in lay novel sets of rules
in metrics and measures .Many author provide stable metrics, which take all possibilities and situation in the account .There are many
ontologies which are emerge from an existing ones. So we need a stable and most elaborative method for ontology analysis. An
Incredible deals of endeavour are being invested in applying semantic web ontologies to create together agreeable and consistent
vocabularies to explain vicinity information from different supplies. We can apply this technology to various areas where there are
thousands of members are related to connected to each other with their same properties .Examples of interconnected RDF dataset
such as FOAF..As ontologies grow in size and number, so it is very important to maintain the ontologies. The maintaining of
ontologies are done by calculating the complexity of ontologies .Many researchers defines many methods. Experiences of the software
engineering field also suggested that there are many correlation between software complexity and quality (such as reusability and
maintainability) [6].

[1]As we believe that is relation between complexity and quality too. More complex ontologies are difficult to maintain. So we have
evaluate the ontologies Because evaluation help to find that ontologies which automatically creating files from different sources that
are heterogeneous in nature and create a multiple and duplicate instances which leads to decrease the usefulness of ontologies and
create a complex structure[2] .

VOLUME 4, ISSUE 6, NOV/2017 28 http://iaetsdjaras.org/


IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

Architecture of federated ontologies of the semantic web


5 Layer architecture of federated ontologies for semantic web.
It is adopted from federated ontologies [9].

Figure 1: Architecture of federated Ontologies

The architecture is extension of standard 3 layer architecture ANSI/SPARC with 2 additional layers. The main layers are [9].

1. Local ontology: Specialized for application.


2. Normalized ontology: Converted form local ontology into a common model.
3. Export Ontology: Describe only relevant part from the normalized ontology.
4. One merged ontology: Global ontology obtain from composition of two or more export schema.
5. Distinct application in the upper layer of ontologies.
Building of ontology can be done in two way:
1. Start from the initial [10].
2. Extend the previous ontology [10].
In both cases, techniques for evaluating the outcome ontology are necessary [2]. Such methods could not only be useful during the
ontology engineering process, they can be useful to an end-user who needs to find the most suitable ontology among a set of
ontologies.

II. LITERATURE REVIEW


In this section we have discussed the analysis of complexity and design of ontologies .Complexity is made up of two Latin words Coms
(means: together) and word plea (means: woven). Complexity is commonly used to signify something with many elements in which
those components have interaction with every one-of-a-kind in more than one methods, pinnacle in a better order of emergence more
than the sum of its elements. The Software complexity is primarily based on famous software metrics, this could be in all likelihood to
lessen the time spent and cost estimation in the checking out segment of the software development lifestyles cycle (SDLC), that could
best be used after software coding is completed.

VOLUME 4, ISSUE 6, NOV/2017 29 http://iaetsdjaras.org/


IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

Examples are like coupling metrics, cyclamate complexities, and the CK object-oriented design metrics. This definition is consistent
with the usage of ontology as set-of-concept-definitions, however more desired. And it is clearly a specific feel of the phrase than its use
in philosophy. An ontology is an explicit specification of a conceptualization [11].

In current decade many metrics were developed to analyse the ontologies properties and its design. In 2004 [12] kang et al suggest
that we could use the entropy based metrics for measuring the complexity of ontology. The main feature of this metrics is to structural
complexity. In 2005 [13] Yao et al describe the 3 matrices Root class, leaf class, depth of inheritance. The metrics are collected by using
different parsers which are used owl ontology. In 2007 [14] Vrandecic and Sure They use technique is based on two notions, first
normalization of an ontology, and 2nd stable metric. Normalization consists of the 5 steps (i) name anonymous instructions, (ii) call
nameless individuals, (iii) materialize the subsumption hierarchy and unify names, (iv) propagate instances to deepest feasible elegance
or assets within the hierarchy, and
(v) Normalize belongings.

Stable metrics are metric which take the open world presumption properly into account, meaning that they may be strong with regards
to possible additions of in addition axioms to the ontology. In 2012 [15] David Snchez proposed a new measure based set which is
made of features of categorization on a concept of ontologies. The results pronounced of measure for endorse a promising accuracy,
improving the correlations said by maximum of different ontology-primarily based techniques, even as minimizing the constraints that
can hamper its applicability each from the computational performance and useful resource-dependency points of view.

In 2005 [16] Andrew Burton-Jones proposed 4 type metric which determine the quality of ontologies based on the different types of
attributes and they are not particularly designed for a specific type of user. They defined metric suites for different type of quality such
as syntactic metric, semantic metric, pragmatic metric, social metric. The ontology should be written in a well-known programming
language. So this is defensible given the growing standardization of ontology languages. The metrics assume that there's are impartial
part of semantics that can be used to evaluate an ontology semantic. To the quantity that the ontology is the simplest regarded
description of some domain, best a part of the metrics will be used.

In 2007 [10] Samir designed a new system for evaluating the ontology in different aspect and prospective .They also refine metric
for evaluation of ontologies by adding some more suite and features like relation diversity , schema deepness ,overall kb metrics and
class specific metrics. According them it provide flexibility to find new ontologies with specific needs .It describe ontology based on
the vocabulary of RDF S and OWL document .This new system used to describe ontologies in a manner that enables the user or
Ontology developer determine the first-rate of an ontology.

In 2006 [3] Derek and Quentin developed a new system called clean onto, it works on the definition for each properties and
concept in the taxonomy. In this approach system, the properties and concept are defined as paths. This have different phases, first is
acquiring description for each node, second is breaking inappropriate nodes links, Thirds is creating the consistent tree. It is a system
which work on the owl file. Clean ONTO gets rid of all inconsistent members of the family from the taxonomic shape, and produces a
steady ontology.

III. THE GRAPH CENTRIC REPRESENTATIONS


[1] When the file size increase and the more the number classes and properties, the more complex of an ontology. The ontologies are
related to each other. We cannot find the complexity by simply counting the number of ontologies, classes and properties. An
ontology are defined according to the directed graph G = {N, R, A}, where N is node of ontologies which is represented as a graph, R
is representing the properties of nodes through which nodes are connected, A is a group edges represents the properties of instances.

{A N X R X N} N is a combination of named and anonymous class, R is a combination of user defined and owl defined properties.
They all are equivalent class of owl (owl: equivalent class). They follow the Translation rules from OWL descriptions to the graph-
centric representation. The graph-centric representation doesnt change the originality of ontology. It is meant to symbolize the
ontology structure with minimum normalization which will constitute the structure of the unique ontology as faithfully as feasible.
The inheritance hierarchy of an ontology can termed as a G = {N, R, A}.They followed the properties of rdfs subclass (rdfs:
subclasses of).

VOLUME 4, ISSUE 6, NOV/2017 30 http://iaetsdjaras.org/


IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

Nodes of top class C N generates an additional edge: E ,if it not already present. The
Black wheel Nodes represent a anonymous classes ,white class represent a named classes .We have to remind that all graph has started
from top level class called T class or thing class.

IV. METRICS FOR ANALYINZING ONTOLOGY


A. Ontology level metrics [1]

Size of vocabulary (SOV): The amount of different meaningful and related word are used in forming the ontology .In the graph G =
{N, R, A} of an ontology, SOV is defined as a relation of names entities and properties: SOV =|Nn| + |Pp| . Where Nn is named
class entities and Pp is properties. It measure the complexity by counting the named entities .Greater complexity leads to large size

Edge node ratio (ENR): In a ontology graph G = {N, R, A}, helps to find how densely the nodes and edges are added and association
between them .Greater the ENR value leads to large complexity and coupling of an ontology. ENR is defined as a
ENR =|A|/|N|. A represents the edge and N represent the Nodes.

Tree impurity (TIP): It helps in measure the inheritance level hierarchy. It is defined as a TIP = |A|+|N|+1.
Where A, N represent the rdfs: subclassOf edges and number of nodes in an ontology graph. A tree with n nodes constantly has n 1
edges, consequently TIP =0. The more TIP leads to greater an ontologys inheritance hierarchy changes from a tree, and the extra
complexity of an ontology.
Number of children: Noc is a metric which is used to find the number of children or subclasses which are associated with the parent
class. It is same as described by chidamber and kemerer. It means that more classes are inter related with each other .If we change on
class, then it will disturb the whole integrity of graph.
Depth of inheritance (DIT): DIT can be measured by the maximum length from parent node to the root node of a tree. It can
calculate the longest path from given node to the root node of graph in ontology inheritance. For accessing the nodes we have to
traverse the graph in top down approach and we have to start with owl: Thing. If there cycle is detected then traversal is to be terminate.

VOLUME 4, ISSUE 6, NOV/2017 31 http://iaetsdjaras.org/


IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

B. Schema Metrics [2] .

Schema metrics based on the design of ontology .The design of ontology based on the richness, width, depth, and inheritance.

Relationship Richness: Relationship richness defined as the proportion of the vector quantity of non-inherited relationship (P) and
sum of vector quantity of total number of inherited relationships (H) and non-inherited relationship (P).

Formula of RR is = |P|/|H|+|P.
|
Inheritance Richness: Inheritance richness defined as the relationship between the ontology tree nodes in terms of properties
(inheritance) between the parent nodes. Low value of IR indicate that the ontology are specified in very detailed manner. This
contained a information very detailed way.

Formula of IR is = |H|/|C| whereas the |C| is number of class in ontology graph.

Attribute Richness = AR can be defined as the number of attributes present in ontology class .If it contain number attributes then
we can concluded that it contain more knowledge about the specific knowledge about the content.

Formula of ARR = |arr|/|C|.

It is the ratio of number of attribute to the number of classes.

C. Knowledge Based Metric.[10]

The way data is positioned inside an ontology is likewise a totally critical measure of ontology quality because it could imply the
effectiveness of the ontology layout and the amount of actual-international information represented by using the ontology. Instance
metrics encompass metrics that describe the KB (Knowledgebase) as a whole, and metrics that describe the way every schema
magnificence is being utilized within the KB.

Class Richness: Class richness is the ratio or value of the number of non - empty classes divided by the total number of classes in the
defined ontology schema.

CR = |C`|/|C|.
This metric is associated with how instances are allotted across the training. The range of training which have instances within the KB is
compared with the total wide variety of instructions, giving a trendy idea of ways properly the KB utilizes the knowledge modeled by
using the schema instructions. Thus, if the KB has a totally low Class Richness, then the KB does not have information that exemplifies
all the magnificence expertise that exists inside the schema.

Class Connectivity: This metric gave a representation about the connectivity of classes in the ontology which totally depends on KB.
With the help of connectivity classes we can determine the nature and working of the ontologies.

Formula of Connectivity = |NIREL (Ci)|.

Cohesion: This metric can be used to signify the life of such cases wherein the KB has more than one linked issue (one being the
correct situation where all instances are related to each other), indicating areas that need more times with the intention to enable
instances from one connect component to connect with instances in other linked additives. Total number of connected component in
knowledge bases are represented as cohesion.

D. Clean onto.[3]

Clean onto is a definition based approach for determining the ontology. It create consistent relation based ontology structure .It is
alternative for the Word net .It create the path between the parent and child node, it uses different algorithm from word net .It have 3
phases evaluation of ontology.

VOLUME 4, ISSUE 6, NOV/2017 32 http://iaetsdjaras.org/


IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

1. Acquiring descriptions for each concept: Firstly it gather information on unclean ontology .This provide path from the named
concept for ontologies.
2. Breaking inappropriate links: Inconsistent link are must be removed. As all the factors of the determine route are found in the
child path, we remember this link to be regular.
3. Creating a consistent tree: After removing redundancy we make a new tree. We decide where to comprise the second sub tree by
figuring out in which to place its root node.

V. ANALYSIS OF QUALITY MEASURING TECHNIQUES


We have gone through the literature from various methods as explained in section II to IV .We find various techniques like ontology
level metrics, scheme metrics, and knowledge metrics & clean onto. Based on the available method and literature we have found that
there are some differences between them, we analysed the some factors which is explained in Table I.

TABLE I. COMPARITIVE ANALYSIS

Sr. No. Methods of Key Uses Preceding/Succeeding Focus on


Evaluation parameters stages
1 Ontology Graph based Business application After ontology Graphical
level ontologies Modeling representation
Metric and Owl
2 Scheme Metric based Evaluation of other After ontology Graphical or
metric ontologies(Business Modeling class level
purpose) ontology
3 Knowledge Metric based Use to describe the After ontology Graphical or
metric measure the real Modeling class level
ontologies(Business ontology
purpose)
4 Cleanonto Step based Evaluate the After ontology Graphical
consistency and Modeling ontology
grow new ontology

VI. CONCLUSION AND FUTURE SCOPE


Various methods are available in the industry .They are efficient in different aspect for ontology evaluation .We have to create a
system which can accommodate the feature of this type of system. Because there are large ontologies present in the industries and
these all are implemented on certain type of examples, so we need that system that evaluate all kind of ontologies. We are looking
forward in this direction to develop uniform system for all kind of ontologies.

REFERENCES

[1] Hongyu Zhang , Yuan-Fang Li , Hee Beng Kuan Tan , Meassuring the design complexity of ontologies .
[2] Samir tartir, I. Budak Arpinar and Amit P seth, ontological evaluation and validation
[3] Derek Sleeman, Clean Onto:evaluation taxonomic realtionship in ontologies.
[4] Brickley, D., Guha, R.V. (Eds.), 2004. Resource Description Framework (RDF) Schema Specification 1.0, (Feb. 2004).
<http://www.w3.org/TR/rdf-schema/>.

VOLUME 4, ISSUE 6, NOV/2017 33 http://iaetsdjaras.org/


IAETSD JOURNAL FOR ADVANCED RESEARCH IN APPLIED SCIENCES ISSN NO: 2394-8442

[5] Horrocks, I., Patel-Schneider, P.F., van Harmelen, F., 2003. From SHIQ and RDF to OWL: the making of a web ontology language. Journal of
Web Semantics 1 (1).
[6] Li, H.F., Cheung, W.K., 1987. An empirical study of software metrics. IEEE Transactions on Software Engineering 13 (6), 697708.
[7] Sheth&Larsen ,Ontology Merging for Federated Ontologies on the Semantic Web.
[8] The Semantic Web T. Berners-Lee, J. Hendler, and O.Lassila. Scientific American 284 (5): 34-43 (May 2001) .
[9] Sheth&Larsen,1990, A bottleneck for federated ontologies in the Semantic Web .
[10] Samir,Budak&Micheal ,Onta Qa: Meteric based ontology quality analysis.
[11] Grubber 1993, A translation approach to portable ontology specifications knowledge accuqusition.
[12] Kang . D ,Xu.B, Lu.j ,A complexity measure for ontologybased on Uml.
[13] Yao, H., Orme, A.M., Etzkorn, L , Cohesion matrics for ontology design and application journal of computer science.
[14] Vrandecic , D., Sure, Y., How to design the better ontogy metrics.
[15] David Snchez1, Montserrat Batet, David Isern, Aida Valls, Ontology-based semantic similarity: a new feature-based approach.
[16] Andrew Burton-Jones, A semiotic metrics for accessing the ontology.

VOLUME 4, ISSUE 6, NOV/2017 34 http://iaetsdjaras.org/

Anda mungkin juga menyukai