INCREMENTAL ONTOLOGY INTEGRATION

Thomas Heer, Daniel Retkowitz

Department of Computer Science 3, RWTH Aachen University, Ahornstr. 55, 52074 Aachen, Germany

Bodo Kraft

AMB Generali Informatik Services GmbH, Anton-Kurze-Allee 16, 52064 Aachen, Germany

Keywords:

Knowledge Management, Ontology Engineering, Information Integration Tools, Human Factors.

Abstract:

In many areas of computer science ontologies become more and more important. The use of ontologies for

domain modeling often brings up the issue of ontology integration. The task of merging several ontologies,

covering speciﬁc subdomains, into one uniﬁed ontology has to be solved. Many approaches for ontology

integration aim at automating the process of ontology alignment. However, a complete automation is not

feasible, and user interaction is always required. Nevertheless, most ontology integration tools offer only very

limited support for the interactive part of the integration process. In this paper, we present a novel approach for

the interactive integration of ontologies. The result of the ontology integration is incrementally updated after

each deﬁnition of a correspondence between ontology elements. The user is guided through the ontologies to

be integrated. By restricting the possible user actions, the integrity of all deﬁned correspondences is ensured by

the tool we developed. We evaluated our tool by integrating different regulations concerning building design.

1 INTRODUCTION

Our approach to ontology integration has been devel-

oped in the context of the ConDes research project

(Kraft, 2007). In this project we have developed new

concepts for software tools to support the conceptual

design phase in building design. Thereby a knowledge-

based approach has been followed. The relevant termi-

nology is deﬁned in several domain-speciﬁc ontologies.

Based on theses ontologies restrictions for the concep-

tual design of a building can be speciﬁed. Because

of this context, the example ontologies in this paper

come from the domain of building engineering, but

our approach for interactive ontology integration is

applicable to many other domains.

There is a broad ﬁeld of different types of

structures that are all subsumed by the term ontol-

ogy (G

omez-P

erez et al., 2004). Ontologies can be

simple vocabularies, i. e. lists of terms, which denote

the entities of a certain domain. If a generalization

relation is deﬁned for these terms, one speaks of a

taxonomy. Both of these types are called light-weight

ontologies. Light-weight ontologies deﬁne concepts,

classiﬁcations of these concepts, properties and rela-

tions. In contrast to that, heavy-weight ontologies com-

prise further semantical information about a domain.

This additional information is speciﬁed by axioms or

constraints. Ontologies can describe concepts on dif-

ferent levels of abstraction. Ontologies, which deﬁne

general concepts, are called upper ontologies, foun-

dation ontologies, or top-level ontologies (Guarino,

1998). Ontologies, which contain knowledge about a

speciﬁc domain, are called domain ontologies.

With regard to ontologies the term integration is

used with several different semantics. Three types of

ontology integration can be distinguished (Pinto et al.,

1999). The ﬁrst type is integration in terms of reuse.

This means constructing a new ontology based on al-

ready existing ontologies, which are incorporated in

the new ontology. A second type is integration in terms

of merging. In this case, two or more ontologies are

uniﬁed into a single ontology by merging correspond-

ing concepts in the original ontologies. The third type

is integration in terms of use. This type of integration

is applied, when applications are built, which are based

on one or more ontologies. In our approach, we use

the term integration in the second sense, i. e. in terms

of merging.

Heer T., Retkowitz D. and Kraft B. (2008).

INCREMENTAL ONTOLOGY INTEGRATION.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - ISAS, pages 13-20

DOI: 10.5220/0001680800130020

 SciTePress

Before merging different ontologies into one uni-

ﬁed ontology, a prior alignment of these is required

(Pinto et al., 1999). Alignment is the process, in which

the relations between the concepts contained in the dif-

ferent ontologies are determined. This is usually done

by the deﬁnition of a mapping between the ontology

elements. This mapping deﬁnes how the source ontolo-

gies have to be merged into one integrated ontology,

so that the resulting ontology contains all the semantic

information of the source ontologies, not more and

not less. One difﬁculty in the alignment of different

ontologies comes from the fact, that the structure of

an ontology is not only determined by the comprised

knowledge, but also by the design decisions made

during its development. Therefore, even ontologies,

which model the same part of a certain domain, may

be structured signiﬁcantly different. This makes the

integration of the ontologies a difﬁcult task.

The paper is structured as follows. First, we give

an overview of related work in section 2. Then, in sec-

tion 3, we will describe how to deﬁne semantic corre-

spondences and how to generate an integrated ontology

from these correspondences. The following section 4

contains the description of our developed integration

algorithm. Next, in section 5, we describe, how the

integrity of the deﬁned semantic correspondences is

ensured. In section 6, we give a short overview of

the tool we developed to implement the integration

approach. Finally, we give a conclusion and an out-

look on further possible developments at the end of

the paper.

2 RELATED WORK

Ontologies and ontology integration are still emerg-

ing topics in the ﬁeld of computer sciences. Many

approaches for the use and integration of ontologies

have been proposed in research.

In (Wache et al., 2001) different techniques for

the alignment of ontologies are described. These are

manual deﬁnition of correspondences, use of linguistic

heuristics, top-level grounding and the use of seman-

tic correspondences. These techniques are not exclu-

sive, but rather complement each other. The ﬁrst tech-

nique requires a knowledge engineer, developing an

ontology, to manually deﬁne certain correspondences

between the concepts of the ontologies to integrate.

These correspondences mainly have the semantics of

equivalence, but are not restricted to 1:1 relations. In

the second method, heuristics are applied to ﬁnd corre-

spondences automatically based on linguistic features

of the terms, representing the concepts. The method

of top-level grounding requires a common top-level

ontology for all ontologies to be integrated. This top-

level ontology is then used to identify related concepts,

and to use this information as a basis for the integra-

tion. Finally, semantic correspondences can be deﬁned.

In this method, different types of semantic relations

are used to relate the concepts of the ontologies to

integrate. This way, not only equivalence relations,

but also relations with other semantics can be deﬁned.

In our approach, we use the techniques of top-level

grounding and semantic correspondences.

In (Kalfoglou and Schorlemmer, 2003) and (Eu-

zenat, 2004) surveys over existing approaches to on-

tology alignment are presented. Both works give an

overview over theoretical frameworks and several cur-

rent research projects. The surveyed works range from

formal over heuristic approaches to approaches, which

use machine learning to automate the process of ontol-

ogy alignment. However, most of the presented works

more or less neglect the issues involved with the inter-

active part of the integration process. In the following

we present two examples of related works, which use

heuristics for the alignment of ontologies.

One alternative to align ontologies is, to consider

lexical similarities between the terms, which represent

the deﬁned concepts. Such a lexical integration ap-

proach is implemented by the tool Chimaera (McGui-

ness et al., 2000). Chimaera is an environment, which

can be used for merging and testing ontologies. When

integrating ontologies, Chimaera generates lists of

suggestions for equivalent terms from the ontologies.

These suggestions are based on lexical similarity mea-

sures. Besides that, Chimaera can identify parts of

the class hierarchy, which probably need to be reorga-

nized. These parts are identiﬁed by means of heuristic

strategies. Since Chimaera uses heuristics based on

lexical analysis, the identiﬁed similarities may contain

mismatches. Thus, it is necessary that the user veriﬁes

all suggestions made by the tool. However, Chimaera

does not propose any solutions in case of conﬂicts,

which may arise during the integration process.

In (Noy and Musen, 2000), an algorithm for semi-

automatic merging and alignment of ontologies called

PROMPT is presented. This algorithm realizes a

semi-automatic integration of ontologies. The Anchor-

PROMPT algorithm (Noy and Musen, 2001) is an

extension to PROMPT. It is used to generate sugges-

tions, which are not only based on linguistic similarity,

but also on structural properties of the ontologies. In

the ﬁrst step, PROMPT generates suggestions for cor-

respondences between the classes of the ontologies, to

be integrated. These initial suggestions are based on

linguistic similarities of the class names and the struc-

ture of the ontologies. The latter is analyzed by the

Anchor-PROMPT algorithm. In the next step, the user

ICEIS 2008 - International Conference on Enterprise Information Systems

selects for each suggestion an operation to perform or

deﬁnes a different operation manually. PROMPT then

automatically performs the selected operation and ap-

plies additional modiﬁcations to the merged ontology,

if required. Subsequently, the list of suggestions is

updated and a list of conﬂicts, which resulted from the

previous operation, is generated. After this, the proce-

dure is executed again, until no more operations have

to be performed, and all suggestions are processed.

In (Hakimpour and Geppert, 2001), ontologies are

used to integrate database schemata. To perform the in-

tegration, the schemata are augmented by correspond-

ing ontologies that deﬁne the schema semantics. These

ontologies are then integrated to a global ontology

from which a uniﬁed schema can be derived. To in-

tegrate the ontologies, similarity relations between

schema concepts are deﬁned. In (Hakimpour and Gep-

pert, 2001) the same four types of similarity relations

are used as in our ontology integration approach. It is

described, how the resulting integrated ontology can

be derived from the source ontologies and the deﬁned

correspondences. There are different suggestions, how

to deﬁne the similarity relations between ontology el-

ements, neither of which is discussed in detail. One

suggestion is, to provide common references by using

a higher level ontology. Other possibilities are, to use

thesauruses, experts familiar with both ontologies, or a

hybrid semiautomatic method. However, no concepts

are proposed, how an expert could be supported in

deﬁning the similarity relations, which includes ﬁnd-

ing corresponding elements and choosing the right

relation types. Nothing is said about, how to ensure

the integrity of the deﬁned correspondences, and how

to take the effects of deﬁned correspondences on the

integration result into account during the alignment of

the ontologies.

3 SEMANTIC

CORRESPONDENCES

In our approach of merging lightweight ontologies,

semantic correspondences are used to relate the con-

cepts of different ontologies to each other. Following

an interactive incremental process, a knowledge en-

gineer deﬁnes correspondences between elements of

the ontologies to be integrated. Based on these cor-

respondences an integrated ontology is automatically

generated.

There are four types of correspondences with dif-

ferent semantics: equivalence, overlap, generalization

and disjointness.

The chosen terms for the semantic correspondence

types overlap and disjoint are rooted in the ﬁeld of

Ontology 1

Ontology 2Correspondence Extensions

toilets, restrooms

restrooms

ladies

restrooms

hallways corridors

toilets

kitchens

Figure 1: Corresponding concepts and their extensions.

set theory. In our ontology integration scenario, the

elements of ontologies are terms, structured in a gen-

eralization hierarchy. The terms represent concepts.

Hence, correspondences between ontology elements

relate concepts to each other. A concept deﬁnes a men-

tal collection of objects or circumstances, which have

common attributes. This collection is called the exten-

sion of the concept (Hakimpour and Geppert, 2001).

Extensions are basically sets. Thus, between two ex-

tensions one of the four possible relationships between

sets must hold. The extensions of two concepts can

be equal or disjoint, one can be a subset of the other,

or the two extensions can overlap, i. e. they have a

nonempty intersection, but are neither equal nor in a

set-subset relation. From these four possible relations

between sets the four correspondence types equiva-

lence, disjoint, generalization and overlap are derived.

In ﬁgure 1, examples of corresponding ontology ele-

ments and their extensions are shown. For example the

generalization correspondence from ladies restroom

to restroom implies that the extension of the former

concept is a subset of the extension of the latter con-

cept. Informally speaking, all ladies restrooms are

restrooms. The disjoint correspondence is a special

case insofar as it is not explicitly represented by an

edge between ontology elements. Whenever no cor-

respondence of one of the other three types is deﬁned

between two concepts, they are implicitly deﬁned as

disjoint.

The deﬁned correspondences between ontology el-

ements allow for the automatic generation of a merged

ontology. In our approach an arbitrary number of

source ontologies can be merged into one ontology.

To illustrate this, we consider here the result of the

integration example from section 4. In ﬁgure 3 c)

cutouts of three ontologies from the domain of build-

ing construction are shown along with the resulting

merged ontology. Several correspondences are deﬁned

between the elements of the source ontologies, e. g.

the terms toilet and restroom are deﬁned as equivalent.

The terms dining room and living room are deﬁned as

overlapping, since there are rooms that have the func-

tionality of both, like e. g. a family room. Therefore

family room is deﬁned as a specialization of dining

INCREMENTAL ONTOLOGY INTEGRATION

room.

During the execution of the integration procedure,

the knowledge engineer selects elements of the merged

ontology to deﬁne correspondences, but the correspon-

dences are established between the elements of the

source ontologies or correspondences, which generate

the selected elements of the merged ontology.

The merged ontology is not generated at once, after

all correspondences have been deﬁned. It is incremen-

tally updated throughout the execution of the integra-

tion algorithm, which is described in the following

section.

4 INTEGRATION ALGORITHM

In our integration approach an arbitrary number of

ontologies can be merged into one ontology, which

contains all the semantic information of the source on-

tologies. In the ﬁrst step two ontologies are integrated.

After that, all remaining ontologies are integrated one

by one into the merged ontology, where each ontol-

ogy is integrated with the current intermediate result.

Except for the choice of terms, which represent the

concepts in the merged ontology, the integration result

is independent of the order, in which the source on-

tologies are integrated. This is guaranteed, because the

correspondences deﬁned throughout the integration

process are established between elements of the origi-

nal source ontologies. The source ontologies remain

unchanged in the knowledge base, and the merged on-

tology can be generated from the source ontologies

and the deﬁned correspondences at any time.

Our ontology integration approach provides so-

lutions to the aforementioned problems. It aims at

providing tool support for the interactive integration

of ontologies. The effects of deﬁned correspondences

are on the one hand taken into account by the fact,

that the knowledge engineer always works with the

current intermediate result. On the other hand, deﬁned

correspondences restrict the possibilities for deﬁning

new correspondences.

The problems of ﬁnding corresponding ontology

elements and deﬁning the correspondences in the right

order are addressed by two aspects of our integration

algorithm. First, a restrictive traversing order of the

merged ontology is enforced. And second, those on-

tology elements, for which correspondences can be

deﬁned at a certain point in the integration processes,

are restricted to a manageable number. This way the

knowledge engineer is guided through the merged on-

tology, and his attention is focused to a relatively small

part of the possibly large ontology.

Our integration approach relies on the assumption,

Figure 2: Highlighting of ontology elements.

that all ontologies use a common top-level ontology.

When integrating two ontologies, one can assume that

the roots are equivalent concepts. Thus, in a ﬁrst step a

top-level grounding of the two ontologies is performed.

After the deﬁnition of equivalence correspondences

between the roots of the ontologies, a ﬁrst version

of the merged ontology is generated. In this merged

ontology the corresponding ontologies are combined

by unifying their roots. After the top-level grounding

an adapted breadth-ﬁrst traversing is performed. The

traversing is steered by the deﬁned correspondences.

At each point during the integration of two ontolo-

gies some elements are highlighted. The knowledge

engineer is only allowed to deﬁne correspondences

between these highlighted elements.

In ﬁgure 2 the highlighting of ontology elements

depending on the type of a previously deﬁned cor-

respondence is shown. For example in ﬁgure 2 a)

an equivalence correspondence has been established

between the ontology elements A and 1. At a later

point in the integration procedure the knowledge engi-

neer is asked to deﬁne correspondences between the

highlighted elements B, C, 2 and 3. In ﬁgure 2 b) a

generalization correspondence between A and 1 has

been established. The ontology element A is deﬁned

to be a specialization of 1. Hence, in a later step the

relationships between A and the specializations of 1,

namely 2 and 3, have to be clariﬁed. The case of an

overlap correspondence constitutes a special case. An

overlap correspondence provides the least information

about the relationships of the specializations of the

linked objects. Thus, the highlighting of ontology ele-

ments is conducted in three steps. In a ﬁrst step, one

of the corresponding elements and the specializations

of the other are highlighted, like it is shown in ﬁgure

2 c). In this step it is not allowed, to deﬁne equivalence

correspondences between A and either 2 or 3, because

ICEIS 2008 - International Conference on Enterprise Information Systems

this would imply, that A is a generalization of 1. For

the same reason it is furthermore not allowed to deﬁne

a generalization correspondence with A as source and

one of the specializations of 1 as target. In a second

step the highlighting of the ontology elements is the

other way round as in the ﬁrst step, while the same

restrictions apply. This situation is depicted in ﬁgure

2 d). While in ﬁgure 2 c) the case of an overlap corre-

spondence without generation (indicated by the dashed

arrows) of an ontology element is shown, in ﬁgure 2

d) the element for the intersection is generated. This

element is not highlighted in steps one and two, be-

cause its relationships to the overlapping concepts are

already deﬁned. Finally, in the third step all special-

izations of the overlapping concepts are highlighted,

including a potentially generated element, to give the

knowledge engineer the opportunity to clarify their

relationships. This situation is not depicted in ﬁgure 2.

Figure 3 shows some steps of the integration al-

gorithm by example. Figure 3 a) shows the situation

directly after the top-level grounding of ontology 1

and ontology 2, as explained earlier. The root con-

cepts of the ontologies are linked by an equivalence

correspondence. Hence, the generated merged ontol-

ogy depicted in ﬁgure 3 a) on the right contains only

one element room and all specializations of room from

the two ontologies are children of this root element.

The elements sanitary room, toilet and dining room are

highlighted. In the following, the knowledge engineer

deﬁnes a generalization correspondence from toilet to

sanitary room. This is the only correspondence, he

deﬁnes for the highlighted elements, and thus the algo-

rithm proceeds. Because of the deﬁned generalization

correspondence, in ﬁgure 3 b) the ontology elements

restroom and toilet are highlighted. The knowledge

engineer deﬁnes an equivalence correspondence be-

tween these elements. In this case, the generalization

correspondence is modiﬁed, so that it references the

equivalence correspondence instead of the ontology

element toilet. This is depicted in ﬁgure 3 c). Figure 3

c) shows the ﬁnal situation, in which a third ontology

has been integrated with the other two ontologies. The

resulting merged ontology contains all the semantic

information of the source ontologies.

5 CORRESPONDENCE

INTEGRITY

Many tools for the integration of ontologies generate

suggestions for possible semantic correspondences be-

tween ontology elements, but provide no assistance

for choosing the right correspondences in the right

order. In our view the main functionality of a tool,

which provides support for the alignment of ontolo-

gies, should be, to ensure the integrity of user-deﬁned

semantic correspondences. The integrity is ensured,

if no conﬂict between deﬁned correspondences exists

regarding their semantics.

In our ontology integration approach, the integrity

of deﬁned correspondences is ensured by several

means. Some of these have already been described

in section 4. The breadth-ﬁrst traversing order en-

sures, that correspondences between general concepts

of the ontologies are deﬁned before correspondences

between their specializations. The former restrict the

possibilities for deﬁning the latter.

The incremental integration approach reduces the

possibility of deﬁning inconsistent correspondences,

because changes like the uniﬁcation of equivalent ele-

ments are immediately performed on the merged on-

tology. The highlighting of ontology elements and the

restrictions, which hold for deﬁning correspondences

between them, also contribute to ensuring the integrity

of deﬁned correspondences, as described in section 4.

These restrictions can be seen as static restrictions, as

they do not depend on previously deﬁned correspon-

dences between the highlighted elements and their

generalizations.

However, when deﬁning correspondences between

the highlighted ontology elements, all previously de-

ﬁned correspondences between the highlighted ele-

ments, between their generalizations, and between the

former and the latter have to be taken into account.

Thereby, we consider the type of the correspondence

to be deﬁned as well as the types of previously deﬁned

correspondences. Restrictions, which arise through

this consideration, are called dynamic restrictions. To

determine all dynamic restrictions for a certain corre-

spondence type, we examined the extensions of those

concepts that should be linked by the new correspon-

dence. In ﬁgure 4 for each correspondence type one

example case is shown, in which the deﬁnition of a cor-

respondence is prohibited. The restrictions result from

the relationships of the concepts’ extensions. Crossed

out correspondences mean, that in the deﬁned case

there must not be a correspondence of the given type,

e. g. in ﬁgure 4 b) the pattern includes all cases, where

there is no generalization correspondence from S to X.

The example in ﬁgure 4 b) is a case in which no

generalization correspondence can be established be-

tween the ontology elements G and S, where G should

be deﬁned as the more general concept and S as its

specialization. The ontology element G is a specializa-

tion of another ontology element X, while S is not. On

the right side, the possible relations between the exten-

sions of G, S and X are shown. The extensions of X and

S can be disjoint, they can overlap, or the extension of

INCREMENTAL ONTOLOGY INTEGRATION

Ontology 1 Ontology 2 Ontology 3 Merged ontology

dining room

room

sanitary room

restroom

room

toilet

g dining roomsanitary room

restroom

room

living room

family room

dining room o

room

sanitary room

restroom

room

toilet

living room

family room

dining roomsanitary room

restroom

room

dining room

room

sanitary room

restroom

room

toilet

dining roomsanitary room

restroom

room

toilet

Figure 3: Example steps of the integration procedure.

X can be a subset of the extension of S. However, in

any case it is impossible, that the extension of S is a

subset of the extension of G. Hence, a generalization

correspondence from S to G is not allowed.

The fewest restrictions apply for the deﬁnition of

an overlap correspondence, because it provides the

least information about the relation of the linked con-

cepts, and is thus compatible with most other corre-

spondences.

Whenever a knowledge engineer is going to de-

ﬁne a new correspondence, all dynamic and static re-

strictions are checked. If some restriction would be

violated by establishing the correspondence, then the

according user action is prohibited by our tool and the

user is informed about the reason for the denial.

Our integration algorithm can be extended by us-

ing heuristics for ﬁnding correspondences between the

highlighted ontology elements. This could be linguis-

tic similarity measures like in (McGuiness et al., 2000)

or heuristics, which take the structure of the ontology

graph into account, like in (Noy and Musen, 2001). In

this way, it could be possible, to achieve a high degree

of automation of the integration process. Correspon-

dences between highlighted ontology elements would

be automatically generated, if their similarity measure

would succeed a certain threshold, and their deﬁnition

would not violate any restrictions. User interaction

would only be required in undecidable cases. We did

not follow this approach, because our main focus lied

on the support for interactive ontology integration.

6 TOOL SUPPORT

We implemented our approach by developing a graph-

based tool for ontology integration. We used the

graph rewriting system PROGRES (Sch

urr et al., 1997)

to specify transformations on a so-called host graph,

which contains the representations of the source on-

tologies, the correspondences and the resulting merged

ontology. The application logic, which constitutes

the core part of the integration tool, was generated

from this speciﬁcation. The graphical user interface

of the integration tool was realized by means of the

UPGRADE framework (B

ohlen et al., 2002).

Our visual graph-based tool provides an abstrac-

tion of the internal data structure and provides a user-

friendly and problem-adequate representation. We

evaluated the applicability of our approach and the efﬁ-

ciency of our integration tool, by merging several large

ontologies from the domain of building design. The

merged ontologies contained the relevant concepts for

the deﬁnition of knowledge for the conceptual design

of the university hospital in Aachen.

7 CONCLUSIONS

In this paper we presented a novel approach for interac-

tive ontology integration. Several different ontologies

can be merged into one. The ontologies are integrated

one by one, while the structure of the resulting merged

ICEIS 2008 - International Conference on Enterprise Information Systems

Equivalence correspondence between 1 and 2 not allowed in the following case:

1 X 2 1

Generalization correspondence from S to G not allowed in the following case:

G X S G

G X

Overlap correspondence between 1 and 2 not allowed in the following case:

1 X2

Ontology elements Possible relations of extensions

o1 2X

G SX

1 SX

Figure 4: Examples for dynamic restrictions for correspondences.

ontology is independent of the order, in which they are

integrated. The alignment of the ontologies relies on

the deﬁnition of semantic correspondences between

their elements. These correspondences are manually

deﬁned by a knowledge engineer. The knowledge en-

gineer is supported by our graph-based tool in many

ways. Alignment and merging steps alternate through-

out the integration process. The intermediate result

of the integration is immediately updated after each

deﬁnition of a correspondence. That way, the effects

of deﬁned correspondences on the integration result

become directly visible. The user is guided through the

merged ontology, and his attention is focused on small

parts of the ontology, where he has to deﬁne new cor-

respondences. In this way, the common problems of

ﬁnding corresponding ontology elements and deﬁning

the correspondences in the right order are substantially

reduced. The integrity of all deﬁned correspondences

is ensured, because actions, which would violate it, are

prohibited by the tool. Thereby, all restrictions, which

arise through previously deﬁned correspondences, are

taken into account. We identiﬁed static and dynamic

restrictions, which may prohibit the deﬁnition of cer-

tain correspondences, and motivated these restrictions

by looking at the extensions of corresponding con-

cepts. The integration algorithm can be extended by

using heuristics for generating suggestions for corre-

spondences. Thereby, a high degree of automation of

the integration process could be achieved. The com-

bination of the two approaches – the calculation of

suggestions for correspondences using heuristics on

the one hand, and the restriction of possible correspon-

dences on the other hand – would probably enable the

tool, to make reliable estimations about the correct

correspondences between ontology elements. So far,

our focus lied on the support for interactive ontology

integration. Nevertheless, it would be a promising ap-

proach, to combine our concepts with approaches for

automatic ontology integration.

REFERENCES

ohlen, B., J

ager, D., Schleicher, A., and Westfechtel, B.

(2002). UPGRADE: Building Interactive Tools for Vi-

sual Languages. In Callaos, N., Zheng, B., and Kader-

ali, F., editors, Proc. of the 6th World Multiconference

on Systemics, Cybernetics, and Informatics (SCI 2002),

pages 17–22, New York. TPA Publishing.

Euzenat, J. (2004). State of the Art on Ontology Alignment.

Technical Report D2.2.3, Knowledge Web Consortium.

omez-P

erez, A., Fern

andez-L

opez, M., and Corcho, O.

(2004). Ontological Engineering: With Examples from

the Areas of Knowledge Management, e-Commerce

and the Semantic Web. Springer.

omez-P

erez, A., Gruninger, M., Stuckenschmidt, H., and

Uschold, M., editors (2001). Proc. of the IJCAI–01

Workshop on Ontologies and Information Sharing, Or-

lando, Florida. AAAI Press.

Guarino, N. (1998). Formal Ontology and Information Sys-

tems. In Guarino, N., editor, Proc. of the 1st Intl. Conf.

on Formal Ontology in Information Systems (FOIS’98),

pages 3–15, Amsterdam. IOS Press.

Hakimpour, F. and Geppert, A. (2001). Resolving Semantic

Heterogeneity in Schema Integration: an Ontology

Based Approach. In Proc. of the 2nd Intl. Conf. on

Formal Ontology in Information Systems, volume 2001,

pages 297–308, New York, USA. ACM Press.

Kalfoglou, Y. and Schorlemmer, M. (2003). Ontology Map-

ping: The State of the Art. The Knowledge Engineering

Review, 18(1):1–31.

Klein, M. (2001). Combining and Relating Ontologies: An

Analysis of Problems and Solutions. In (G

omez-P

erez

et al., 2001), pages 53–62.

Kraft, B. (2007). Semantische Unterst

utzung des

konzeptuellen Geb

audeentwurfs. Dissertation, RWTH

Aachen University, Aachen.

INCREMENTAL ONTOLOGY INTEGRATION

McGuiness, D. L., Fikes, R., Rice, J., and Wilder, S. (2000).

An Environment for Merging and Testing Large On-

tologies. In Giunchiglia, F. and Selman, B., editors,

Proc. of the 17th Intl. Conf. on Principles of Knowl-

edge Representation and Reasoning (KR-2000), pages

483–493, San Francisco, USA. Morgan Kaufmann Pub-

lishers.

Noy, N. F. and Musen, M. A. (2000). PROMPT: Algorithm

and Tool for Automated Ontology Merging and Align-

ment. In Engelmore, R. and Hirsh, H., editors, Proc.

of the 12th Conf. on Innovative Applications of Artiﬁ-

cial Intelligence (IAAI-2000), pages 450–455, Orlando,

Florida. AAAI Press.

Noy, N. F. and Musen, M. A. (2001). Anchor-PROMPT:

Using Non-Local Context for Semantic Matching. In

omez-P

erez et al., 2001), pages 63–70.

Pinto, H. S., G

omez-P

erez, A., and Martins, J. P. (1999).

Some Issues on Ontology Integration. In Benjamins,

V. R., Chandrasekaran, B., G

omez-P

erez, A., Guarino,

N., and Uschold, M., editors, Proc. of the IJCAI–99

Workshop on Ontologies and Problem-Solving Meth-

ods: Lessons Learned and Future Trends (KRR5), vol-

ume 18 of CEUR Workshop Proceedings, pages 7/1–

7/12, Aachen. Departement of Computer Science 5,

RWTH Aachen.

Sch

urr, A., Winter, A., and Z

undorf, A. (1997). The PRO-

GRES approach: Language and environment. In Ehrig,

H., Engels, G., Kreowski, H.-J., and Rozenberg, G., ed-

itors, Handbook on Graph Grammars and Computing

by Graph Transformation: Applications, Languages,

and Tools, volume 2, pages 487–550. World Scientiﬁc.

Wache, H., V

ogele, T., Visser, U., Stuckenschmidt, H.,

Schuster, G., Neumann, H., and H

ubner, S. (2001).

Ontology-Based Integration of Information – A Survey

of Existing Approaches. In (G

omez-P

erez et al., 2001),

pages 108–117.

ICEIS 2008 - International Conference on Enterprise Information Systems