Novel Approach For Effective Ontology Alignment Computer Science Essay

Published:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

For more precise and better information retrieval on the semantic web, where meaningful but sometimes irrelevant information is retrieved, but using ontology mapping there could be an improvement in getting more relevant information. Ontology alignment is the process of finding the similarity between the concepts in a heterogeneous environment. This paper presents an approach for ontology mapping. Two different ontologies of a particular domain are considered and the concepts which are similar to each other from both of the ontology, are retrieved i.e. ontology mapping. Also the similarity is being calculated if the two concepts are not matched even by expanding the term. The conceptual analysis of the technique shows that the results obtained through the proposed approach provides the semantic terms of the same domain.

Keywords: Ontology, Ontology Mapping, Ontology Alignment

1 Introduction

Nowadays, the internet has become the main source of retrieving the information. All the information is stored on the web in structured, semi-structured or unstructured form. So, when any of the particular information is searched over the internet, it returns a large number of relevant as well as irrelevant documents. To overcome the existing problem in the traditional web, the concept of semantic web came into existence. Semantic web deals with the knowledge base (KB) where the data is stored in a structured form. Ontology is a special type of KB. In general, ontology is defined as the branch of science that deals with all the concepts of a specific domain and set up the connection between them i.e., classes, subclasses and other entities of a specific domain and the relationships that bind them. As ontology is domain specific and everyone has its own perception when creating ontology so it results in heterogeneity. To have a more precise information retrieval the ontologies are needed to be merged or fused. In semantic web the description adds to the concepts by using the data from the ontology. Integration of ontology is a very difficult job because of the heterogeneity in them. Although there are many approaches for ontology integration but only a few are able to handle them effectively. The problem of heterogeneity can be overcome by: i) Finding the set of correspondence and ii) Alignment based on the application requirement.

The main aim of ontology mapping is to find the concepts that represent the meaningful connection within similar concepts of different ontology. In our approach we are checking the similarity of the concepts by taking different measures. i.e. string similarity, semantic similarity and similarities between the concepts by calculating the weight of each concept w.r.t their features.

1.1 Organization of the Paper

The remaining part of the paper is organized as follows: Section 2 introduces the Terminology; related work regarding the matching approaches is included in Section 3; Section 4 constructs overall system architecture and the proposed algorithm along with the examples; section 5 provides the conclusion.

2 Terminology

In this section we are presenting the basic terms used in our approach.

2.1 Ontology

Ontology can be defined as:

O = {N, SN, OP, DP, L}

Where N is set of nodes with their corresponding sub-nodes, denoted by SN.

OP is an object property that establishes the relation between the concepts.

Addition of data of a concept related to the particular context is data property, denoted by DP.

L is leaf node i.e. A node that does not have further nodes.

As the Fig.1 describes, Ontology is a type of database where the data is stored in the hierarchical or structured form. The relation between the concepts is established in ontology. Ontology is represented in web ontology language (OWL) and is created at the top of RDF (Resource description framework). RDF provides a complete description of a resource. It consists of three types of object types: resource, relationship and value.

<rdf:RDF>

<owl:Class rdf:about="http://www.onto.com/auto.owl/Audi"/>

<owl:Class rdf:about="http://www.onto.com/auto.owl/Color">

<rdfs:subClassOf rdf:resource="http://www.onto.com/auto.owl/Audi"/>

</owl:Class>

<owl:Class rdf:about="http://www.onto.com/auto.owl/Speed">

<rdfs:subClassOf rdf:resource="http://www.onto.com/auto.owl/Audi"/>

</owl:Class>

<owl:Class rdf:about="http://www.onto.com/auto.owl/Transmission">

<rdfs:subClassOf rdf:resource="http://www.onto.com/auto.owl/Audi"/>

</owl:Class>

</rdf:RDF>

Fig. 1. Example of an ontology describing classes and sub-classes

2.2 Mapping

It is a process of finding the semantic association between the concepts used in different ontology. Ontology mapping results in either fusion of ontology or alignment of a set of corresponding terms.

3 Related Work

Ontology mapping has been an active research area in recent years. Patrick et al [2] created a technique named SAMBO and applied it in biomedical ontologies. It is used for merging and mapping ontologies. The output is 1:1 alignment between terms and relations. If the terms are in a similar position with respect to, part-off or is-a then it is considered that the concepts are similar with each other. As soon as the matching has been done, the system can merge the similar concepts. Trong et al [11] presented an approach named as Anchor-prior which enhances the quality of combining the ontology by minimizing the computational complexity. It starts from finding the two matched concepts and then their neighbors i.e. Superclass, subclass, siblings etc. are analyzed step by step. Similarity is calculated among Priorly Matachable Concept (PMC) in two different ontologies. In this algorithm pair of terms in different ontologies which are similar to each other w.r.t. structure and description are considered firstly. The total computational complexity calculated for this approach is O (n*log n).

The divide and conquer approach was also used in ontology mapping by Hu et al [3]. The input can be either RDFS or OWL. Falcon operates in three stages 1) division of ontology 2) construct cluster 3) find corresponding concepts. Agglomerative clustering algorithm is used for finding the similarity and then the clusters are constructed accordingly. Clusters of both the ontologies are matched and the clusters having high similarities are selected. QOM (Quick Ontology Mapping) [9] uses dynamic programming for increasing run-time efficiency. The main aim of this approach was to increase the efficiency and effectiveness. It can provide the ability to exchange and use information between ontologies when they cannot be integrated. ASMOV [6] is an approach for mapping the ontology automatically by calculating the similarity and validating the semantics. The output for the given OWL ontologies is many-to-many i.e. n: m alignment between the ontologies. The weights are adjusted automatically to the features of the ontology. In the first phase lexical, external, internal and individual similarities are calculated and then the alignment algorithm is applied that validate the alignment at the end of each looping. Pruning process is used to resolve all the semantic issues like crisscross mapping. Memetic algorithm [12] was proposed for the automatic alignment between two ontologies by adding the local search on an existing genetic algorithm and searching is performed by using the local information for taking the decision for the next move.

Although the previous approaches for ontology mapping are applicable but did not focus on GUI (graphical user interface) and also there are limitations in taking input in different related languages like RDFS, SKOS, XML, N3 etc. and also some of the existing approaches sometimes proved to be less efficient in case of large ontologies. In this paper we are focusing on overcoming these limitations to some extent.

4 Proposed System

Our proposed algorithm results in a set of matched concepts in different ontologies. In this approach, first of all we divided all the nodes into a single unit and then find the corresponding nodes one by one. It works in three steps:

Case 1: Check for Syntactic similarity

For calculating syntactic similarity, the Levenshtein algorithm is used. It calculates the minimum number of characters required to change the one string into another. For example, in the given two ontologies it is applied for the two subclasses Scooter and Scooter. The relation between them is found as equivalence relation.

i.e. Scooter = Scooter

Case 2: Check for semantic similarity

It is checked when the concepts are not syntactically similar. WordNet is used to extract the synonyms of a specific concept. After expanding the concept, lavenshtein algorithm is used again to find the similarity. The class Motorcycle is expanded as:

Motorcycle € {bike, motor vehicle, automotive vehicle} and the similar term is,

Motorcycle ≡ bike

Case 3: Finding similarity by calculating the weight

In this case, certain weight is assigned to each of the features and then the similarity is calculated. The weight (W) for each candidate is calculated and compare during mapping.

W= i

n

where F is feature and n is total features.

1 if wc1≈ wc2 and n1 ≈ n2

sim(C1 , C2) =

Otherwise

4.1 Architecture

According to the proposed architecture (see Fig. 2), ontologies of a specific domain are mapped and a set of aligned terms are returned. Everyone has its own perception in creating ontology which results in heterogeneity in ontology knowledge base. So to overcome the problem of heterogeneity in ontology knowledge base, ontology mapping is used, which results in a new ontology knowledge base consisting of the terms from each ontology knowledge base, which would help to retrieve the data more efficiently and accurately. The semantic annotation is performed on the result to modify the results and a semantic information retrieval system is created through which the query from the user is processed through user interface. A closer view of ontology mapping is shown in Fig. 3.

Document Set

Aligned Ontology

Semantic IR

Semantic annotation

U s e r I n t e r f a c e

Ontology B

Ontology A

Mapping

Fig.2. Overall System Architecture

Ontology A

{ k1,k2 …ki}

ki

tj

Syntactic

Similarity

Semantic

Similarity

Calculate

Weight

Similarity

Aligned

Terms (ki , tj)

Compare

Weight

Ontology B

{ t1,t2 ……ti}

Fig.3. Functional structure of ontology mapping

4.2 Algorithm

In the proposed algorithm, ontology is divided into a single term and then each term is mapped with the term of other ontology. Mapping could be syntactic, semantic or comparing weights. Result consists of a set of similar terms, known as ontology alignment. Step 1to 5th shows the division of ontology into a single term and each term of level (hierarchy) is checked step-by-step (Step 6- 8). Step 9 to 10th return aligned terms if the terms are matched (syntactically). If the terms are not syntactically similar then expand the term using dictionary and return similar if any of the synonym is matched as shown in step 11 to 14th. Else calculate the weight using the property of the term and compare it with the term of other ontology. Return similar if weights of the term of the ontologies are similar(Step 15-22). At last repeat from step 6-22 for the next level(step 23).

Input: Ontology A and Ontology B

Output: Set of similar terms

n= no. of nodes (n to m)

level= no. of levels in ontology A

if n%2 == 0

then n ← n/2

else n ← (n+1)/2

while(n>1)

n <- n/2

l <- 0

while(l<level)

while(n<=m)

If A[n]==B[k] // k=1,2,3………., j-I, j

Return(Similar)

A[l]= n+1,n+2….m

else n= n1,n2, … , nj

If (ni == B[k]) //i= 1,2,…..,j

Return(Similar)

Check for all the properties P

P= {P1, P2 ,…………, Pn}

For each P

Pi = weight

For each Concept C

C= (P1 + P2 +…… + Pn) /n // n is total properties in C

If (Ci ≈ Cj) AND (ni ≈ nj)

Return (Similar)

l=l+1

4.3 Demonstration of Algorithm with Example

Example 1

As shown in Fig. 4, in ontology A, a term "Sedan" and a term "Saloon" in ontology B. For finding the semantically same term, mapping (Fig. 5) is done between the ontologies. The resulted aligned terms are shown in Fig. 6.

Fig. 4. Ontology A and B constructed using Protégé

≡≡

≈

≡≡

≡≡

=

Automobile

TwoWheeler

FourWheeler

Motorcycle

Scooter

Hatchback

StationWagon

PassengerVehicle

EstateCar

CompactCar

Jeep

Truck

GoodsVehicle

Lorry

Vehicle

Sedan

Car

Bike

Scooter

Saloon

≡≡

≈

Fig. 5. Mapping of Ontology A and B

Similarity

Ontology A

Ontology B

Semantic (≡)

Motorcycle

Bike

Syntactic (=)

Scooter

Scooter

Compare weight (≈)

Sedan

Saloon

Compare weight (≈)

Hatchback

CompactCar

Semantic (≡)

StationWagon

EstateCar

Semantic (≡)

Truck

Lorry

Fig. 6. Resulted Aligned Terms

Example 2

We are taking two data sets as ontology A and ontology B. Ontology A consists of some individuals like BMW7Series, AstonMartinDBS etc. with some features like speed, fuel, body type etc. and likewise ontology B. Certain value is assigned to each of the feature and then the weight for each concept is calculated by using the formula:

W= i

n

and the terms having weight equal or approximately equal are returned as similar term.

Ontology A

Features

BMW7Series

AstonMartinDBS

RevaClasse

Truck

hasMake

BMW

AstonMartin

Reva

Tata

hasSpeed

10-15km/h

10-15km/h

80 km/h

Above 20km/h

hasBody

Luxury

Luxury

Hybrid

-

hasFuel

Petrol

Petrol

Electric

Diesel

hasTransmission

Automatic

Manual

Automatic

-

hasRoof

-

Sunroof

-

-

Ontology B

Features

BMW3Series

VolvoS8

RevaA/C

hasBrand

BMW

Volvo

Reva

hasSpeed

10-15km/h

15-20km/h

80 km/h

hasBody

Luxury

Luxury

Hybrid

hasFuel

Petrol

Petrol

Electric

hasTransmission

Automatic

Automatic

Automatic

hasRoof

Moonroof

-

-

Feature

Weight

BMW

0.062

AstonMartin

0.035

Reva

0.097

10-15 km/h

0.020

80 km/h

0.001

Luxury

0.026

Hybrid

0.088

Petrol

0.009

Electric

0.010

Automatic

0.038

Manual

0.099

Sunroof

0.047

Moonroof

0.070

Volvo

0.130

15-20 km/h

0.409

Tata

0.004

Diesel

0.015

Above 20km/h

0.350Assign some weight to each of the features as:

Calculation of weight

Weight of the features is calculated on the basis of their weighted properties. Each property is being assigned a weight. For example, P is a set of properties denoted by P1, P2, ….. , Pn

Weight (W) = (P1 + P2 +, ….. , + Pn)

n

BMW7Series (0.062+0.020+0.026+0.009+0.038) / 5 = 0.031

Aston Martin (0.035+0.020+0.026+0.009+0.009+0.047) / 6 = 0.0228

RevaClasse (0.097+0.001+0.088+0.010+0.038) / 5 = 0.0468

BMW3Series (0.062+0.020+0.026+0.009+0.038+0.070) / 6 = 0.0375

This value is approximately equal to the value of BMW7Series. So, both of these are similar.

RevaA/C = 0.097+0.001+0.088+0.010+0.038) / 5 = 0.0468

Similar with RevaClasse as the values are similar.

VolvoS8 = (0.130+0.409+0.026+0.009+0.038) /5 = 0.1224

Not similar with any of the calculated weight.

Truck = ( 0.004+0.015+0.350) /3 = 0.123

The calculated weights are compared and the terms having calculated value approximately equal are returned as similar.

Fig. 7. Graph showing results

Fig.7 shows the comparisons of the terms with respect to their calculated weights. The results calculated using the proposed approach is more accurate as compared to other approaches.

6 Conclusion

The aim of our proposed approach is to solve the ontology alignment problem. To achieve the goal of retrieving precise and relevant result, the ontology mapping is done. On the basis of experiment done it is concluded that the proposed algorithm is enhancing the accuracy and efficiency of the result. It worked in two phases: i) mapping and ii) alignment. Use of this technique in some semantic web application is the future work of this approach. Using this and other approaches the searching could be modified.

Writing Services

Essay Writing
Service

Find out how the very best essay writing service can help you accomplish more and achieve higher marks today.

Assignment Writing Service

From complicated assignments to tricky tasks, our experts can tackle virtually any question thrown at them.

Dissertation Writing Service

A dissertation (also known as a thesis or research project) is probably the most important piece of work for any student! From full dissertations to individual chapters, we’re on hand to support you.

Coursework Writing Service

Our expert qualified writers can help you get your coursework right first time, every time.

Dissertation Proposal Service

The first step to completing a dissertation is to create a proposal that talks about what you wish to do. Our experts can design suitable methodologies - perfect to help you get started with a dissertation.

Report Writing
Service

Reports for any audience. Perfectly structured, professionally written, and tailored to suit your exact requirements.

Essay Skeleton Answer Service

If you’re just looking for some help to get started on an essay, our outline service provides you with a perfect essay plan.

Marking & Proofreading Service

Not sure if your work is hitting the mark? Struggling to get feedback from your lecturer? Our premium marking service was created just for you - get the feedback you deserve now.

Exam Revision
Service

Exams can be one of the most stressful experiences you’ll ever have! Revision is key, and we’re here to help. With custom created revision notes and exam answers, you’ll never feel underprepared again.