CROSS DOMAIN KNOWLEDGE VERIFICATION

Verifying Knowledge In Foundation Ontology Based Domain Ontologies

Najam Akber Anjum, Jenny Harding and Bob Young

Wolfson School of Mechanical and Manufacturing Engineering, Loughboruogh University

LE11 3TU, Loughborough, U.K.

Keywords: Knowledge sharing, Knowledge verification, Foundation ontologies, Ontology matching.

Abstract: Knowledge verification refers to the process of making sure that the knowledge shared between knowledge

bases of two parties is correctly understood on both sides. Domain ontologies developed out of a foundation

ontology have a potential to improve the knowledge verification methods. This can be done by following

concepts in domain ontologies to their origin and constituent conceptualisations in the foundation ontology.

This is possible when matching ontologies belonging to two different domains but developed out of a single

foundation ontology. Along with the concepts, a prescribed way of using these concepts by domain

ontology builders also needs to be included in the foundation ontology. This prescribed way can exist in the

form of an ontology of constraints which governs and shapes the building of domain ontologies according to

the needs of the verification system and thus makes them more interoperable.

1 INTRODUCTION

Knowledge verification systems are needed to

ensure the correct understanding of concepts and

related knowledge among parties involved in the

knowledge sharing activity. This paper proposes a

way of verifying the authenticity of knowledge and

concepts across different domain ontologies

developed out of a single foundation ontology. It

first gives a brief review of the literature on existing

domain ontologies and ontology matching tools, a

discussion about the proposed ontology matching

methodology follows and conclusion and future

work is presented in the end.

2 EXISTING RESEARCH

2.1 Foundation and Domain Ontologies

To make knowledge bases more shareable and

expandable, instead of building them from scratch, it

is more apt to develop them out of a single agreed

upon foundation or standard (Neches et al, 1991).

Foundation ontologies provide the basis for this

standard. They make the expansion and integration

of knowledge bases easier. This is because if two

system builders build their knowledge bases on a

common ontology, the system will share a common

structure, and it will be easier to merge and share the

knowledge bases (Swartout et al, 1997).

Some of the existing foundation ontologies

include Standard Upper Ontology – SUO (Niles &

Pease, 2001), Suggested Upper Merged Ontology –

SUMO (Niles & Pease, 2001), DOLCE (Gangemi et

al, 2002), WordNet (Deng et al, 2009), and Cyc

Ontology (Matuszek et al, 2006).

Foundation ontologies like these may help in

reducing semantic heterogeneity by restricting

domain ontology builders to match their own

conceptualisations against a common foundation, so

that all communication is done according to the

constraints derived from the ontology (Schorlemmer

& Kalfoglou, 2005). Domain ontologies on the other

hand provide a set of terms for describing some

domain (Swartout et al, 1997) and they can be

thought of as taxonomies of relevant objects within

that domain. Example of domains may include

aerospace, biology, manufacturing, arts etc.

2.2 Use of Foundation Ontologies for

Ontology Matching

Foundation or upper ontology are being used to

match concepts in two independently developed

ontologies. The idea is to first match two ontologies

with an upper ontology and then matching these two

ontologies using the similarities existing between

them and the upper ontology. The LOM tool

339

Akber Anjum N., Harding J. and Young B..

CROSS DOMAIN KNOWLEDGE VERIFICATION - Verifying Knowledge In Foundation Ontology Based Domain Ontologies.

DOI: 10.5220/0003065203390342

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 339-342

ISBN: 978-989-8425-29-4

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

developed by Li (2004) uses WordNet, SUMO and

MILO to match two ontologies. Aleksovski et al

(2006) use DICE ontology as background

knowledge to match two flat unstructured lists of

concepts. Mascardi et al (2007) present an algorithm

which uses upper ontologies to align two

heterogeneous ontologies. They also experiment

with OpenCyc, SUMO-OWL and DOLCE as

semantic bridges to match ontologies (Mascardi et

al, 2010). All of these cases deal with situations

where independently developed heterogeneous

ontologies are matched by using an upper ontology

through the process of semantic bridging. These

bridges are built when individual ontologies are

matched with the upper ontology. The research

presented in this paper, however, proposes to build

these bridges during the ontology development

process by the domain ontology builders so as to

provide the knowledge verification system with

clues to establish concept similarity. The verification

system in this way uses the foundation ontology as a

dictionary to interpret ontology based knowledge

across diverse domains as has been proposed in the

research literature (Mascardi et al, 2007, 2010). To

achieve this there needs to be provided a set of core

concepts along with their prescribed use. This

‘prescribed use’ needs to be there to ensure that a

trace is left of every domain concept or a

combination of concepts to the foundation ontology

counterparts and it is this dimension of the proposed

verification framework here which distinguishes it

from the rest of the research work described in this

section.

2.3 Knowledge Verification

When formalized knowledge is shared between

different domains, prevention of its subjective

interpretation becomes necessary. This process of

authentication of the interpretation, here, is referred

to as knowledge verification. The description which

endorses the sense in which the term verification is

used here is the one given by Gupta (1993) where he

mentions that knowledge verification involves the

checking of completeness, consistency and

correctness of knowledge. For this to happen during

the cross domain knowledge sharing between

ontology based knowledge bases, similarities first

need to be established between the two ontologies.

This leads to the need of an ontology matching and

mediation system. The techniques these systems use

are discussed next.

2.4 Ontology Mediation Techniques

The most crucial stage in the ontology mediation

task is the similarity finding. Different mediation

tools use different strategies and algorithms to

achieve this purpose. These matching algorithms can

be divided into four types (Aleksovski et al, 2006).

I. Terminological methods which are based on

lexical matching of ontological concepts,

II. Instance-based methods where the lexical

similarity of instances is compared in two

ontologies,

III. Structural methods where the positions of

different concepts in the structure of two

ontologies are used to find matches and

IV. Semantic methods which use some additional

logic to discover similarities.

Each of these methods or a combination of them

is employed by the ontology mapping and matching

tools to overcome mismatches that exist in

independently developed heterogeneous ontologies.

Two main types of mismatches that may come up

when matching ontologies are explication and

conceptualization mismatches (Visser et al, 1993). A

closer look at the ontology mediation tools which

use these similarity finding methods reveals that

most of these tools are just able to overcome

explication mismatches and even fewer do this

automatically (Anjum et al, 2010).

The matching and verification technique

explained in this paper partly resembles the fourth

type of similarity finding methods explained here

and it capable of overcoming not only the

explication but also the conceptualization

mismatches. This method includes the use of

semantic information in the form of connections

between the domain ontology concepts and their

foundation ontology counterparts. These connections

are established during the ontology building process.

The challenge is therefore to make sure that some

standard connections, like inheritance, are put in

place during the ontology building stage for the

verification system to make use of. This can be

achieved through a verification meta ontology which

is explained in the sections to come.

3 FOUNDATION ONTOLOGY

MAPPING FRAMEWORK

This system is proposed specifically for domain

ontologies formed by using the concepts from a

foundation ontology. Figure 1 shows the schematic

of this framework. The framework consists of three

modules. The ‘inheritance identifier’, ‘domain

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

340

concepts identifier’ and ‘concept matcher’. The

whole process of matching consists of six steps. The

sequence of these steps is indicated in the figure

through circled numbers. These steps are explained

below:

Foundation

Concepts

InheritanceIdentifier ConceptsMatcher

Foundation Ontology

CoreConceptsOntology

VMO

DomainConcepts

Identifier

Inheritance

Queries

Inherita nce

Queries

Foundationandrelevant

domainconcepts

Replies

Equivalentfoundat ion concepts

Veri f i er

DesignDomainOntology

ManufacturingDomainOntology

Domainand relevant

foundationconcepts

Figure 1: The Verification framework.

1- The process initiates with the generation of a

query from the inheritance identifier. This query

explores the foundation concepts behind a

concept in the design domain ontology. This

domain concept is the one which is required to

be matched with a similar concept in the

manufacturing domain ontology.

2- The inheritance identifier receives replies in the

form of relevant foundation conceptualisations.

3- These foundation conceptualisations and

relevant design concepts are then sent to the

‘concept matcher’ and to the ‘domain concept

identifier’ at the same time.

4- The ‘domain concept identifier’ then generates a

query to explore the concepts in the domain

manufacturing ontology which possesses the

same foundation concept inheritance.

5- The domain concept identifier then receives the

replies in the form of domain concepts having

the same foundation concepts as sent by the

inheritance identifier.

6- These domain concepts along with their related

foundation concepts are then sent to the concept

matcher and concepts with similar foundation

inheritance are declared as similar.

3.1 Verification Meta Ontology

The purpose of a verification meta ontology (VMO)

is to police the use of concepts from the foundation

ontology. This can be done by loading the

foundation ontology and the VMO along with the

newly developed domain ontologies in the ontology

editing environment. In this setting the VMO

scrutinizes all the concepts for traceability to

foundation ontology and if a lag is found the domain

ontology builders are notified. For example, when

the concept of hole is referred to in the foundation

ontology, as shown in figure 2, it can be given any

name as long as it has enough semantic information

about its origin in the foundation ontology and that

is what the VMO controls.

hole

bolt_hole

joining

feature_1

Source ontology Target ontology

Foundation ontology

Foundation

Level

Domain

Level

Figure 2: Foundation ontology subsumptions in domain

ontology.

This VMO may consist of a few classes but

predominantly a set of rules governing the use of

concepts from the foundation. Through these rules,

VMO performs the process of semantic enrichment

of domain concepts. It makes sure that there is

enough evidence or traceability of concepts formed

in the domain ontology in order for them to be

tracked back in the foundation ontology. The VMO

needs to be built by the verification system builders

keeping in view the contents and structure of the

foundation and core concepts ontology.

3.2 A Possible Scenario

The example of a disc is taken here which is to be

modelled in a domain ontology by using core

concepts from the foundation ontology. Figure 2

shows the difference in interpretations of features of

the same disc in design and manufacturing and how

the same disc is modelled differently in two domain

ontologies. The foundation for the concepts used in

domain ontologies, however, are same for both

design and manufacturing. The dotted lines show

how the concept of ‘disc_edge’ feature is inherited

from the foundation concept of ‘rim’. Similarly

‘diaphragm’ is connected to the foundation concept

of ‘web’. Establishment of these inheritance

relations can be made compulsory, through the

verification meta ontology. This might be needed

when a new concept is introduced in a domain

ontology. Some examples of these inheritance

relations can be:

same_as

different_from

is_a_type_of

It is through these relations and other compulsory

attributes that the verification system proposed

above will be able to track the identity of a concept

in the foundation ontology and thus will verify the

knowledge shared. The example given here is the

CROSS DOMAIN KNOWLEDGE VERIFICATION - Verifying Knowledge In Foundation Ontology Based Domain

Ontologies

341

Figure 3: A possible scenario.

simplest possible case. More complex cases may

include a totally different interpretation of features

of a disc in design and manufacturing domains.

4 CONCLUSIONS

It can be inferred from the above propositions that a

foundation ontologies need to come with a set of

core concepts, a verification meta ontology and a

knowledge verification system which interprets

concepts across different domains by using the

VMO rules. Domain ontologies developed by using

this toolkit will be interoperable no matter what

terminologies and combination of concepts they use

to model entities. Knowledge associated to these

models would therefore be shareable and verified.

The most important thing for this verification

system to work is, therefore, the information and

knowledge capturing. This is because it is that stage

where the domain ontology concepts are

semantically enriched for the verification system to

work. The dynamic nature of this technique makes it

much better than just mapping the similar concepts

manually in two ontologies. The technique is

dynamic because it allows the ontology builders to

make changes and modifications during the life time

of the ontologies without caring about its mappings

with other domain ontologies. this is because if the

changes made adhere to the prescriptions of the

verification meta ontology they are easily

interpretable by any ontology which is built on the

same rules and uses concepts from the same

foundation ontology.

REFERENCES

Anjum, N., A., Harding, J., A., Young, B. and Case, K.,

2010. Gap Analysis of Ontology Mapping Tools and

Techniques. In: K. Poppelwell, J. Harding, R. Poler

and R. Chalmeta, eds, Enterprise Interoperability IV.

1st edn. UK: Springer, pp. 303-312.

Aleksovski, Z., Klein, M.C.A., Ten Kate, W., and

Harmelen, F. Van, 2006, “Matching Unstructured

Vocabularies Using a Background Ontology”, Proc.

Int. Conf. Knowledge Eng. and Knowledge

Management (EKAW ’06)

Aleksovski, Z., Ten Kate, W., and Harmelen, F. Van,

2006, “Exploiting the Structure of Background

Knowledge Used in Ontology Matching”, in Shvaiko,

P., Euzenat, J., Noy, N., Stuckenschmidt, H.,

Benjamins, R., and Uschold, M. eds, 2006, “Proc. Int’l

Workshop Ontology Matching (OM-2006)”

Deng, J., Dong, W., Socher, R., Li, L.-., Li, K. and Fei-

Fei, L., 2009. ImageNet: A Large-Scale Hierarchical

Image Database, CVPR09, 2009, .

Ehrig, M. and Staab, S., 2004. Qom – Quick Ontology

Mapping. pp. 683-697. Proceedings of the Third

International Semantic Web Conference, Springer

Gangemi, A., Guarino, N., Masolo, C., Oltramari, A., and

Schneider, L., 2002, “Sweetening ontologies with

Dolce”, In Gómez-Pérez, A., and, Benjamins, V. R.,

ed., 2002, “Proc. of EKAW 2002”, pages 166–181.

Springer,

Gupta, U. G., 1993. Validation and verification of

knowledge-based systems: A survey. Applied

Intelligence, 3(4), pp. 343-363.

Li, J., 2004, “LOM: A Lexicon-Based Ontology Mapping

Tool”; Proc. of the Workshop on Performance Metrics

for Intelligent Systems PerMIS ’04

Mascardi, V., Rosso, P., and Cordi, V., 2007, “Enhancing

Communication inside Multi-Agent Systems—An

Approach Based on Alignment via Upper Ontologies”,

Proc. Int’l Workshop Agents, Web-Services and

Ontologies: Integrated Methodologies

Mascardi, V., Locoro, A., and Rosso, P., 2010,

“Automatic Ontology Matching via Upper Ontologies:

A Systematic Evaluation"; IEEE Transactions on

Knowledge and Data Engineering

Matuszek, C., Cabral, J., Witbrock, M. and DeOliveira, J.,

2006. An Introduction to the Syntax and Content of

Cyc. AAAI Spring Symposium, .

Neches, R., Fikes, R., Finin, T., Gruber, T., Patil, R.,

Senator, T. and Swartout, W. R., 1991. Enabling

technology for knowledge sharing. AI Mag., 12(3), 36-

56.

Niles, Ian and Pease, Adam, 2001. Towards a standard

upper ontology, FOIS '01: Proceedings of the

international conference on Formal Ontology in

Information Systems, 2001, ACM pp2-9.

Schorlemmer, M. and Kalfoglou, Y., 2005. Progressive

ontology alignment for meaning coordination: an

information-theoretic foundation, AAMAS '05:

Proceedings of the fourth international joint

conference on Autonomous agents and multiagent

systems, 2005, ACM pp737-744.

Swartout, B., Ramesh, P., Knight, K. and Russ, T., 1997.

Toward Distributed Use of Large-Scale Ontologies.

AAAI Symposium on Ontological Engineering.

Visser, P. R. S., Jones, D. M., Bench-Capon, T. J. M. and

Shave, M. J. R., 1997, An Analysis of Ontology

Mismatches; Heterogeneity versus Interoperability In

AAAI1997 Spring Symposium on Ontological

Engineering, Stanford, USA.

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

342