ONTOLOGY MAPPING BASED ON ASSOCIATION RULE

MINING

C. Tatsiopoulos and B. Boutsinas

University of Patras, Artificial Intelligence Research Center, GR-26500, RIO, Greece

University of Patras, Department of Business Administration, GR-26500, RIO, Greece

Keywords: Ontology Mapping, Interoperability, Association Rule Mining.

Abstract: Ontology mapping is one of the most important processes in ontology engineering. It is imposed by the

decentralized nature of both the WWW and the Semantic Web, where heterogeneous and incompatible

ontologies can be developed by different communities. Ontology mapping can be used to establish efficient

information sharing by determining correspondences among such ontologies. The ontology mapping

techniques presented in the literature are based on syntactic and/or semantic heuristics. In almost all of

them, user intervention is required. In this paper, we present a new ontology mapping technique which,

given two input ontologies, is able to map concepts in one ontology onto those in the other, without any user

intervention. It is based on association rule mining applied to the concept hierarchies of the input ontologies.

We also present experimental results that demonstrate the accuracy of the proposed technique.

1 INTRODUCTION

Ontology engineering, i.e. designing, developing,

maintaining and sharing ontologies, is an emerging

knowledge engineering process. It allows the

information organization into taxonomies of

concepts, represented by attributes, and relationships

between concepts, represented by IS-A relations,

functions, constraints, etc. Ontologies find

acceptance in numerous applications, e.g.

information retrieval (Pretschner & Gauch, 1999),

document management (Lacher & Groh, 2001),

agent communication (Huhns & Singh, 1997),

finance (Firat, & Madnick, 2001) and e-commerce

(Omelayenko, 2001). However, ontologies are

imposed by the explosive growth of the Semantic

Web, where they are used to describe the semantics

of the data. They are used for conceptually

structuring data and for knowledge sharing.

Ontologies can be designed and developed by

different groups of people with similar interests, i.e.

communities within the so-called information

society, either through knowledge engineering

processes or through automated knowledge

extraction methods.

One of the most important properties of both the

WWW and the Semantic Web is decentralization

((Berners-Lee, 1999), W3C

). Therefore, ontologies

can be designed and developed by different

communities without adopting common standards

for information exchange. On the other hand, the

leverage of synergies of information exchange has

been increased by the deployment of systems for

community interaction support. Many researchers

(e.g. (Lacher & Groh, 2001), (Maedche & Staab,

2000), (Mitra & Wiederhold, 2002), (Stumme &

Maedche, 2001)) argue that common to all systems

ontologies can not be guarantied (see (Wache,

Vögele, Visser, Stuckenschmidt, Schuster, Neumann

& Hübner, 2001) for a survey of such effort),

because it is more efficient if a smaller community is

involved in the process and, in general, communities

can usually not be forced to adopt common

standards. Then an efficient ontology-based

information exchange can be established by solving

the problem of determining correspondences among

different ontologies, i.e. determining the set of

similar, overlapping or unique concepts. This

problem is an instance of the interoperability

problem (e.g. (Park & Ram, 2004)) which concerns

the connection of information systems that are

heterogeneous and incompatible. Recently, this

problem has been a major focus of both the research

and the practitioner communities. Both data and

knowledge engineering has been focused on

identifying correspondences between ontologies and

Tatsiopoulos C. and Boutsinas B. (2009).

ONTOLOGY MAPPING BASED ON ASSOCIATION RULE MINING.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Information Systems Analysis and Speciﬁcation, pages 33-40

DOI: 10.5220/0001862800330040

 SciTePress

between database schemas respectively (e.g. (Choi,

Song & Han, 2006), (Kalfoglou & Schorlemmer,

2003), (Rahm & Bernstein, 2001), (Shvaiko &

Euzenat, 2005)). Ontology heterogeneity and

incompatibility is due to the existence of knowledge

either in different structures or in different

environments with different semantics. Ontology

mapping aims at tackling structural and semantic

heterogeneity and incompatibility by determining

correspondences between elements of disparate

ontologies. Note that structural heterogeneity has

also been addressed to a great extent in the schema

matching literature (Rahm & Bernstein, 2001). A

mapping can be established either directly between

two ontologies (alignment) or indirectly through

mapping them onto a third reference ontology which

both of them share as a common upper model

(articulation). The work of mapping ontologies is

performed mostly by hand, perhaps supported by a

graphical user interface. Of course, performing

ontology mapping manually is an extremely time-

consuming and error-prone process. The ontology

mapping techniques presented in the literature are

usually based on syntactic and/or semantic

heuristics. The latter have been studied in various

scientific fields including machine learning, concept

lattices, formal theories, databases and linguistics. In

almost all of them user intervention is required, thus

they are semi-automated. Usually, when an

automatic decision is not possible, these techniques

suggest possible correspondences, determine

conflicts and propose solutions and actions. Then the

user makes the final selection. In this paper, we

present a new ontology mapping technique, which,

given two input ontologies, is able to map concepts

in one ontology onto those in the other, without any

user feedback. The proposed technique exploits the

structure of the input ontologies, i.e. the concept

hierarchies, to determine the mapping. More

specifically, in the proposed ONARM technique,

ONtology mapping is based on Association Rule

Mining, which extracts association rules from these

concept hierarchies. Association is one of the most

popular data mining tasks. Association rules can be

used to represent frequent patterns in data, in the

form of dependencies among concepts-attributes.

The extracted association rules are considered as

indirectly describing the concept relationships. Note

that, despite the support or the controversy of the

statement that ontology mapping is similar to

database schema matching ((Kalfoglou &

Schorlemmer, 2003), (Noy & Klein, 2002)), the

proposed methodology can be applied to both of

them. In the rest of the paper we first present related

work (Section 2) and then we present the proposed

ontology mapping technique (Section 3). Next, we

present experimental results of testing its accuracy

and efficiency (Section 4). We also discuss its time

complexity (Section 5) and finally we conclude

(Section 6).

2 PREVIOUS APPROACHES

Recently, the number of ontology matching

techniques and systems has increased significantly

(OMO, Ontology Matching Organisation for a

complete information on the topic). Ontology

mapping techniques vary in input and output formats

as well as in modes of user intervention. There has

been little work on the comparative evaluation of

ontology mapping techniques in the literature (e.g.

(Giunchiglia, Yatskevich, Avesani & Shvaiko,

2008), (Kalfoglou & Schorlemmer, 2003), (Kaza &

Chen, 2008)).

There are techniques which simply guide the

user to create the mappings, e.g. PROMPT (Noy &

Musen, 2000), SMART (Noy & Musen, 1999),

PROMPTDIFF (Noy & Musen, 2002), CHIMAERA

(

McGuinness, Fikes , Rice & Wilder, 2000).

There are also semi-automatic techniques in

which the user has to resolve conflicts and

duplicates, FCA-Merge (Stumme & Maedche,

2001), to create mappings for concepts that cannot

be matched, GLUE (Doan, Madhavan , Domingos

&, Halevy, 2002), to validate the matches, ONION

(Mitra & Wiederhold, 2002). Also, some techniques

allow user to suggest matches apart from those

created automatically, e.g. EER-CONCEPTOOL

(Compatangelo & Meisel, 2003).

Moreover, there are techniques which create the

mapping automatically, e.g. CAIMAN (Lacher &

Groh, 2001), IF-Map (Kalfoglou & Schorlemmer,

2002), ITTalks (Prasad, Peng & Finin, 2002),

MAFRA (Maedche, Motik, Silva & Volz, 2002), S-

MATCH (Giunchiglia, Shvaiko & Yatskevich,

2004).

Additionally, there are techniques which are

based on the combination of different matching

processes (e.g. (Aumueller, Do, Massmann & Rahm,

2005), (Hu & Qu, 2008)), which exhibit remarkable

results in term of accuracy (OAEI).

There are also techniques that could potentially

be used in ontology mapping like translators (e.g.

OntoMorph (Chalupksy, 2000)) or integrators (e.g.

Hovy, 1998). Finally, a similar problem is that of

schema matching in databases. However, most

schema matching techniques are not adequate for

ICEIS 2009 - International Conference on Enterprise Information Systems

ontology mapping due to not handling differences in

terminology, due to exhibiting poor results in the

case of little structural similarity, due to absence of

instances, etc.

3 THE PROPOSED TECHNIQUE

The key idea of the proposed technique is to

establish a similarity between two concepts of the

input ontologies, which is based on their location in

the ontology structures. The location of a node, that

represents a concept within an ontology structure,

determines its neighbour concepts. We consider that

the meaning of a concept is also characterized by the

meaning of its neighbour concepts, as the creator of

the ontology indirectly determined by defining the

structure of the ontology.

Note that structural mapping alone is not

sufficient for ontology mapping. The meaning of the

concept is also characterized by a linguistic analysis

of the concept with respect to a large-scale

dictionary like WordNet, to a corpus of documents,

to manual rules, to lexical distances, etc. The

proposed technique accepts both of these sources of

background knowledge in order to establish a

similarity measure. However, the latter is dominated

by the location of a concept within the ontology.

Graph matching techniques could be used in

order to examine the similarity of the location of two

input concepts. Since we concentrate on efficiency,

we rejected such techniques because of their time

complexity (for instance time complexity of graph

isomorphism is exponential). The proposed

technique considers each path of the ontology

structure as a source of background knowledge. It

applies association rule mining in order to determine

the predominant neighbour concepts of an input

concept.

In this paper, we consider association rule

mining that is known as the market basket problem,

where concepts-attributes represent products and the

initial database is a set of customer purchases

(transactions). This particular problem is well-

studied in data mining. We consider association

rules analog to the form “90% of the customers that

purchase product x also purchase product y”

(Boolean association rules) (e.g. (Agrawal, Mannila,

Srikant & Verkamo, 1996), (Brin, Motwani, Ullman

& Tsur, 1997), (Park, Chen & Yu, 1995)).

Formally, an association rule is a rule of the form

X Y, where X,Y named respectively antecedent

and consequent of the rule and X,Y I = {i

…i

}, such that XI Y =

⇒

⊂

∅

and i

, 1 k j is an

item in the transaction database D. The informative

power (named interestingness) of each association

rule is measured by two indexes: the “support” that

measures the proportion of transactions in D

containing both X and Y and the “confidence” that

measures the conditional probability of the

consequent given the antecedent.

≤ ≤

More specifically, the proposed technique

considers each path of the ontology structure as a

transaction. Then, for each input ontology, it applies

association rule mining to the set of its transactions.

We consider that the extracted association rules

determine the predominant neighbour concepts of

every input concept. Thus, the similarity of these

association rules defines the location-based

similarity of the concepts.

Linguistic analysis is also taken into

consideration. However, it is used to increase or

decrease the obtained location-based similarity (see

γ parameter below). In that sense, any heuristic for

linguistic analysis proposed in the literature can be

used. Also, aggregating the results of such heuristics

could also be used, as for example presented in

(Ehrig & Staab, 2004), (Ehrig & Sure, 2004). In this

paper, we adopt a naive such heuristic: we examine

identity of labels of concepts, while we use a

common vocabulary for both ontologies. Obviously,

more advanced heuristics would increase the overall

accuracy.

The proposed technique can be applied to

ontology structures forming a directed acyclic graph.

Thus, it supports multiple inheritance. The required

formal definition of input ontologies contains two

core items shared by most formal definitions of an

ontology in the literature: concepts and a

hierarchical IS-A relation. Thus, we define a core

ontology as: a pair G = (C, r), where C is a set of

concepts and r is a partial order on C, i.e. a binary

relation r

∈

C x C which is reflexive, transitive, and

antisymmetric.

More specifically, the proposed technique

accepts two ontologies as input. Any ontology editor

can be used to create them (we used the Protégé

knowledge-modeling environment). During a first

step the input ontologies are transformed to RDF

and RDFS formats. Obviously, any ontologies pre

described in RDF(s) can be used. Then, the Java –

Jena API is used. Jena is a Java implementation of

an extension to the semantic web by means of a

respective API. This offers the capability of getting

the complete description of the input ontology in

terms of its structural elements (paths, current nodes,

successor nodes, parent nodes, siblings and leaf

ONTOLOGY MAPPING BASED ON ASSOCIATION RULE MINING

nodes). In order to apply the APRIORI association

rule mining algorithm (Agrawal et al.), nodes must

be topologically numbered. Thus, the second step is

to generate a numbered node structure, of the same

structure as the ontology under examination but

numbered with integer numbers that will undergo

Breadth First Search (BFS), such that, integers are

horizontally incremented and assigned and therefore

guaranteeing this way, that any node N

is numbered

with an integer k such that k>m, where m is the

integer which has been assigned to its parent node M

(this relation holds true for any parent and child

nodes within the input ontology). This way, nodes

at deeper levels are mapped with higher number

values.

After numbering the Ontology, the result is a

hash table that includes all the nodes of the ontology

with their respective integers. Then, to extract all

the possible paths, with the objective to quick reach

and examine with priority terminal nodes its paths, a

Depth First Search (DFS) is run, that provides all

possible paths number-named in a list type format.

The methodology has been designed in such a way

that permits multiple inheritance (and therefore has

multiple parents) in the following way and under the

definition:

L(i) is the level of this node i and

Li = max(l1, l2,..ln), where,

l1, l2,…, ln are level numbers to which the node

i belongs simultaneously.

To resolve the above, during the numbering

process, the integers that are assigned to nodes are

non-continuous but retain the necessary property

needed for the APRIORI algorithm that follows,

such that:

for any two nodes I,J, order (J) > order (I), where

node J is parent of node I.

This step involves the extraction of all root-to-

leaf paths available in the ontology schema, by

means of a recursive method. Furthermore, a list of

all leaf nodes is created.

Then, for a predefined set of minimum support

and confidence values, APRIORI algorithm is

applied to both input ontologies (e.g. G

and G

The result is a set of rules of the following type:

1: { [2, 7], 45, 20}

3: { [1, 7, 4], 30, 70}

4: { [1, 3, 6], 45, 20}

5: { [1, 3, 4], 30, 70}

………

where, the integers above denote number-named

nodes of the ontology. Each pair of (c,s) produces

such one respective set of rules R(G

), R(G

Following, the above rule set is back translated and

represented with the original node names, providing

this way R’(G

), R’(G

) of rules.

The ONARM technique generates an [n x m]

“significance matrix” containing the significance in

matching every node of G

=(C

, r

) with every one

of G

=(C

, r

). Note that G

is mapped against G

and not vice versa, considering G

as our reference

ontology. The significance in matching XЄC

YЄC

is calculated based on the support measure of

the association rules having X and Y as left part. For

instance, consider the following rules for X and Y:

X→(B,s1), X→(BC,s2), Y→(B,s3), Y→(AC,s4)

For each of the four pairs of rules, one for X and the

other for Y, the measures K and Kt, indicating the

significance, are computed by the following

procedure:

1. K=0, if |s

– s

| > α, where α a user defined

threshold

2. K ← Average(s

X ,

) * β * w, where

β > 1 if X or Y or both are instances,

β = 1 otherwise,

w = percentage of similarity of right parts

3. Kt = K * γ, where

γ > 1 if X ≡ Y, i.e. the two nodes are

identical after a linguistic analysis

γ = 1 otherwise.

Thus, processing the pair (X→(B,s1),Y→(B,s3))

K=Average(s1,s3)*1*1, processing the pair (X→

(B,s1),Y→(AC,s4)) K=Average(s1,s4)*1*0 and

processing the pair (X→ (BC,s1),Y→(AC,s4))

K=Average(s1,s4)*1*0.5

Then, the matrix is filled as follows:

For every cell (i ,j ) in the [n x m] matrix

a. Calculate

∑

K, for every pair p of

rules, one for i and the other for j

b. Calculate Kt

c. Fill cell (i, j) with Kt

d. Reduce considered cases by using

constraint of maximal selection of Kt,

for subsequent analyses.

Continue

until all ( i,j) cases are filled and reduced.

ONARM, for some minimum support and

confidence (e.g. (s,c)=(25,5)) extracts rules from two

ontologies (e.g. G

) and builts the significance

matrix (e.g. see Table 1). Finally, it provides final

mappings along with their significance, e.g.:

[G2]:ooo50 -> [G3]:ooo50 75.0

In the example of Table 1, nodes of G

are listed

in rows and those of G

are listed in columns of

significance matrix, while cells contain the

significance.

ICEIS 2009 - International Conference on Enterprise Information Systems

Table 1: A portion of the significance matrix.

4 EMPIRICAL RESULTS

For evaluation purposes, a total of 400 combinations

of (s,c) are examined. All 400 cases of possible (s,c)

are summarized in Table 2:

Table 2: Accuracy w.r.t. support and confidence.

What is depicted in the above cumulative

summarized table (only some entries are presented

out of the 400 possible), is the relevance of the final

score used for the mapping, in terms of support and

confidence parameters, that have been run against

and are presented in descending manner, from the

most relevant to the least.

Empirical tests aim at examining the accuracy of

the proposed technique and for this reason the

following experiment has been set up.

Consider three Ontologies G

, G

and G

, in such

a way that G

is directly derived from G

and G

directly derived from G

(see Figures 1-3).

Therefore G

is considered as the Reference

Ontology that we run tests against for G

, G

After applying ONARM to them in a manner of

against G

and G

against G

we obtain the

following summarized results:

a) For the comparison and mapping of the G

and

Ontologies, ONARM found 76 correct

matchings between them. The best cases were

found in the areas where the minimum support

was between 5% - 40% and the minimum

confidence 5% - 65%. Thus, minimum support

and c onfidence values are not critical. The only

Figure 1: Test ontology G

Figure 2: Test ontology G

Figure 3: Test ontology G

ONTOLOGY MAPPING BASED ON ASSOCIATION RULE MINING

requirement is to set low values. Theoretically,

this is true because of the small number of paths

of an ontology.

b) In the last column is presented the number of

matches per case, as non-zero-values

[NonZeroVal].

c) The average, variance and standard deviation

analysis is based upon the score that has been

assigned to each case, as Kt, as described in the

theoretical background section.

d) Cumulative results are presenting, in a more

analytical way, in Table 3 (for G

) and Table 4 (

for G

) for an indicative set of values (s,c):

Table 3: Cumulative results for G

Table 4: Cumulative results for G

Given the above, for G

& G

the success rate is

149 cases out of 400 for the first mapping attempt,

i.e., 38%, while in the comparison of G

& G

the

success rate increases dramatically to 220 out of 400

cases, i.e., 55%,. Analytical results exist for all the

above cases too, providing similar distribution in

respect to support and confidence of the success

cases.

It is important to note here, that the percentages

refer to the average of all cases of s, c. To continue

our investigation, we consider the comparison and

mapping of G

& G

, for s = 5 and c = 5. In this

case, we obtain 11 full matches (FM), 2 zero

matches (ZM) and 0 partial matches (PM) as

follows:

FM(G

) = {ooo50, ooo40, oo175, ooo90, oo160,

oo100, oo170, oo225, oo 195, oo 400, oo500 }

PM(G

) = { }

ZM(G

) = {oo650, oo800}

The ZM(G

) is obtained since our methodology,

puts weight more in the structural elements of nodes,

than to their lectical name. Then, ONARM applies

linguistic analysis to nodes also of the final mapping

process, and adds ZM(G

) to FM(G

). This way,

concluding the process, the precision=1 and recall=1

for the specific mapping of nodes respectively. This

way, ONARM after its lectical mapping phase, finds

100% of correct mappings of G

and G

nodes. Note

that the case of G

& G

is similar.

5 DISCUSSING EFFICIENCY

Considering two input ontologies, the time

complexity of ontology mapping techniques

presented in the literature varies from the order of

O(|G

| * |G

|) (e.g. (Maedche et al.)) to exponential

(e.g. (Stumme & Maedche, 2001), (Giunchiglia, et

al., 2004)).

Recently, there is a concern on the efficiency of

the proposed in the literature ontology mapping

techniques. For instance the work presented in

(Ehrig & Staab, 2004) tries to reduce the search

space by introducing certain strategies to select the

pair of concepts checking for matching. Thus the

time complexity is reduced to

O( (|G

| + |G

|) * log(|G

| + |G

|))

In the proposed ONARM technique, the

extraction of association rules is performed

separately for each ontology, which is O(|G

|*|C

|),

where |C

|) is the number of all candidate itemsets

checked. According to (Agrawal et al.), the

complexity of locating the itemsets of size k is

O(klog(|G

| / k)) (however in random databases there

are only a few large itemsets). In ONARM the

maximum size is d, the depth of the ontology. Thus,

the latter complexity is O(∑

j=1

jlog(|G

|/j) which is

O(d

log(|G

|). Thus, the overall complexity of

extraction is O(d

log(|G

|). The used similarity

matrix is obtained in O(|G

|*|G

|).

6 CONCLUSIONS

We presented an ontology mapping technique,

ONARM, which exploits a structural similarity

measure in order to automatically determine the

ICEIS 2009 - International Conference on Enterprise Information Systems

mapping between two input ontologies. Since

ONARM is based on the structure of ontologies, it

can handle both metadata and instance

heterogeneity. ONARM can easily be included in

systems based on combination of matching

techniques, especially because only a few techniques

exist for only structural similarity ((Ehrig & Staab,

2004), (Euzenat & Valtchev, 2004), (

Hu, Jian, Qu &

Wang,

2005)). Note that ONARM exhibits a low

time complexity with respect to related approaches.

The meaning of the concepts is also taken into

consideration, (step 3 of the procedure calculating K

and Kt), by applying any linguistic analysis. Thus, it

is important to note that input ontologies might have

different label domains for node naming, without

reducing the efficiency of the proposed

methodology.

We plan to continue the evaluation of ONARM

using benchmark ontologies ((

Giunchiglia et al., 2008),

OAEI). Note that early results on specific OAEI-

2008 benchmarks (e.g. 101,102, 201-205, 223) show

almost the highest accuracy.

Also, we are currently working on extending

ONARM in order to automatically update each one

of the input ontologies with respect to the other, by

using a new distance measure (Boutsinas &

Papastergiou, 2008) of the similarity of two concepts

of the same ontology.

ACKNOWLEDGEMENTS

We are grateful to Kostas Sidiropoulos for his help

during the tests.

REFERENCES

Agrawal, R., Mannila, H., Srikant, R., Verkamo, A.I.,

1996. Fast Discovery of Association Rules, in:

(Fayyad, 1996), pp. 307-328.

Aumueller D., Do H.H., Massmann S., Rahm E., 2005.

Schema and ontology matching with COMA++,

Proceedings of SIGMOD (Demonstration).

Berners-Lee, T., 1999. Weaving the Web, Harper.

Boutsinas, B., Papastergiou, T., 2008. On clustering tree

structured data with categorical nature, Pattern

Recognition, 41(12), Elsevier Science, pp. 3613-3623.

Brin, S., Motwani, R., Ullman, J.D., Tsur, S., 1997.

Dynamic Itemset Counting and Implication Rules for

Market Basket Data, in: Proceedings of ACM

SIGMOD Int. Conf. Management of Data, pp. 255-

264.

Chalupksy, H., 2000. OntoMorph: a translation system for

symbolic knowledge, Proceedings of the 17th

International Conference on Knowledge

Representation and Reasoning (KR-2000).

Choi, N., Song, I.Y., Han, 2006. H., A Survey on

Ontology Mapping, SIGMOD Record 35(3), pp. 34-

41.

Compatangelo, E., Meisel, H., 2003. "Reasonable"

support to knowledge sharing through schema

analysis and articulation, Neural Computing &

Applications, - Springer.

Doan, A., Madhavan, J., Domingos, P, Halevy, A., 2002.

Learning to map between ontologies on the semantic

web, Proceedings of the 11th International World

Wide Web Conference (WWW 2002).

Ehrig, M., Staab, S., 2004. QOM— Quick Ontology

Mapping. In: Proceedings of the Third International

Semantic Web Conference, Springer.

Ehrig, M., Sure, Y., 2004. Ontology Mapping - An

Integrated Approach. In Proceedings of ESWS.

Euzenat J., Valtchev P., 2004. Similarity-based Ontology

Alignment in OWL-Lite, Proc. of ECAI, 2004, pp.

333-337.

Fayyad, U.M., Piatetsky-Shapiro, G., Smyth, P., 1996.

Advances in Knowledge Discovery and Data Mining,

AAAI Press/MIT Press.

Firat, A., Madnick, S., 2001. Knowledge Integration to

Ontological Heterogeneity: Challenges from

Financial Information Systems, in 23rd International

Conference on Information Systems (ICIS),

Barcelona, Spain.

Giunchiglia, F., Shvaiko, P., Yatskevich, M., 2004. S-

Match: an Algorithm and an Implementation of

Semantic Matching, In Proceedings of ESWS.

Giunchiglia, F., Yatskevich, M., Avesani, P., Shvaiko, P.,

2008. A Large Scale Dataset for the Evaluation of

Ontology Matching Systems, Knowledge Engineering

Review Journal, (to appear).

Hovy, E., 1998. Combining and Standardizing Large-

Scale, Practical Ontologies for Machine Translation

and Other Uses, proceedings of LREC.

Hu. W., Qu Y., 2008. Falcon-AO: A practical ontology

matching system, Web Semantics: Science, Services

and Agents on the World Wide Web, Volume 6, Issue

3, pp. 237-239.

Hu. W., Jian N., Qu Y., Wang Y., 2005. GMO: a graph

matching for ontologies, Proceedings of K-

CAPWorkshop on Integrating Ontologies, pp. 41-48.

Huhns, M.N., Singh, M.P., 1997. Ontologies for agents,

IEEE Internet Computing 1(6), pp. 81-83.

Kalfoglou, Y., Schorlemmer, M., 2002. Information-

flow-based ontology mapping, Lecture Notes in

Computer Science 2519, pp. 1132-1151.

Kalfoglou, Y. Schorlemmer, M., 2003. Ontology

Mapping: the State of the Art, The Knowledge

Engineering Review 18(1).

Kaza, S., Chen, H., 2008. Evaluating Ontology Mapping

Techniques: An Experiment in Public Safety

Information Sharing. Decision Support Systems

Volume 45, Issue 4, pp. 714-728, Information

Technology and Systems in the Internet-Era.

ONTOLOGY MAPPING BASED ON ASSOCIATION RULE MINING

Lacher, M.S., Groh, G., 2001. Facilitating the exchange

of explicit knowledge through ontology mappings,

Proceedings of the 14th International FLAIRS

Conference.

Maedche, A., Motik, B., Silva, N., Volz, R., 2002.

MAFRA — A MApping FRAmework for Distributed

Ontologies. In: Proceedings of the EKAW 2002.

Maedche, A., Staab, S., 2000. Semi-automatic

engineering of Ontologies from texts, Proceedings of

the 12th International Conference on Software

Engineering and Knowledge Engineering (SEKE

2000), pp. 231-239.

McGuinness D.L., Fikes R., Rice J., Wilder S., 2000. An

Environment for Merging and Testing Large

Ontologies, Proceedings of the Seventh International

Conference on Principles of Knowledge,

Representation and Reasoning (KR2000)

Breckenridge, Colorado.

Mitra, P., Wiederhold, G., 2002. Resolving

Terminological Heterogeneity In Ontologies,

Proceedings of the ECAI'02 workshop on Ontologies

and Semantic Interoperability.

Noy, N.F., Klein, M., 2002. Ontology evolution: not the

same as schema evolution, Also as: Smi-2002–0926,

University of Stanford, Stanford Medical Informatics.

Knowledge and Information Systems (In Press).

Noy, N.F., Musen, M., 1999. SMART: automated support

for ontology merging and alignment, Proceedings of

the 12th Workshop on Knowledge Acquisition,

Modelling and Management (KAW'99).

Noy, N.F., Musen, M., 2000. PROMPT: algorithm and

tool for automated ontology merging and alignment,

Proceedings of the 17th National Conference on

Artificial Intelligence (AAAI'00).

Noy, N.F., Musen, M., 2002. PROMPTDIFF: a fixed-

point algorithm for comparing ontology versions,

Proceedings of the 18th National Conference on

Artificial Intelligence (AAAI'02), pp. 744-751.

Omelayenko, B., 2001. Integration of product ontologies

for B2B marketplaces: a preview. SIGecom Exch.,

2(1).

Ontology Matching Organisation, (OMO)

http://www.ontologymatching.org

Park, J~S., Chen, M.~S., Yu, P.~S., 1995. An Effective

Hash-Based Algorithm for Mining Αssociation rules,

in Proceedings of ACM SIGMOD Int. Conf.

Management of Data, pp. 175-186.

Park, J., Ram, S., 2004. Information Systems

Interoperability: What Lies Beneath? ACM

Transactions on Information Systems 2004; 22(4),

pp. 595-632.

Prasad, S., Peng, Y., Finin, T.,2002. Using Explicit

Information To Map Between Two Ontologies .

Workshop on “Ontologies in Agent Systems”.

Pretschner, A., Gauch, S., 1999. Ontology based

personalized search, Proc. 11th IEEE Intl. Conf. on

Tools with Artificial Intelligence, pp. 391-398.

Rahm, E., Bernstein, P.A., 2001. A Survey of Approaches

to Automatic Schema Matching, VLDB Journal, 10(4).

Shvaiko, P., Euzenat, J., 2005. A Survey of Schema-based

Matching Approaches. Journal on Data Semantics

(JoDS), IV, LNCS 3730, pp. 146—171.

Stumme, G., Maedche, A., 2001. Ontology Merging for

Federated Ontologies on the Semantic Web,

Proceedings of the International Workshop for

Foundations of Models for Information Integration

(FMII- 2001).

Wache, H., Vögele, T., Visser, U., Stuckenschmidt, H.,

Schuster, Neumann, G., H., Hübner, S., 2001.

Ontology-Based Integration of Information - A Survey

of Existing Approaches, in IJCAI Workshop:

Ontologies and Information Sharing, Seattle, WA.

W3C, www.w3.org/DesignIssues/Principles.html.

ICEIS 2009 - International Conference on Enterprise Information Systems