a scalable and efficient query answering for a context and ... · pdf filea scalable and...

Journal of Convergence Information Technology Volume 5, Number 1, February 2010

A Scalable and Efficient Query Answering for a Context and Schema Mediation

Benharzallah Saber *1, Kazar Okba *2

*1,*2

[email protected] of Computer science, University of Mohamed Khider Biskra, 07000 Algeria

, [email protected] doi: 10.4156/jcit.vol5.issue1.3

Abstract

This paper presents a generic, scalable and

efficient query answering for semantic interoperable information systems. We propose a generic multi agent architecture that supports our approach. Moreover, we define a semantic proximity allowing to separate the semantic interoperable information systems in segments allowing encircled efforts of query answering. Another advantage of our approach consists the exploitation of intelligent agents for query reformulation, and using a new technology for the semantic representation. The algorithm reformulates the query by using the schema mediation method for the discovered systems and the context mediation for the other systems.

Keywords

Query answering; Semantic mediation; Multi-agent

systems and OWL DL

1. Introduction

Many organizations face the challenge of interoperating among multiple independently developed database systems to perform critical functions. The cooperation of systems is confronted with many problems of heterogeneities and must take account of the open and dynamic aspect of modern environments. Thus, querying the distributed ontologies that is one major task in semantic interoperable information systems.

Various types of heterogeneity can be met: technical, syntactic, structural and semantic heterogeneity. Several types of semantic conflicts appear: naming conflicts (taxonomic problems and linguistics), values conflicts (units and scales problems ...). Semantic interoperability is becoming more important than before. The solutions for the interoperability of the information systems evolved to the semantic mediation of the systems [1][5].

The rise of the number of information sources implies the increase and the diversification of the

conflicts number, as well as an increase in the time of localization of relevant information; it increases also the time of transmission of the queries towards all these information sources and the response time of the information sources. Therefore, the solutions of semantic interoperability should have an intelligent processor for query answering that allows the adaptation of the changes of the environment and solving the various data conflicts in a dynamic way. Each various existing solutions provide advantages to the detriments of others and each one of them treats just one that part of the data conflicts.

In this paper we propose a generic, scalable and

efficient query answering for semantic interoperable information systems. We propose a generic multi agent architecture supporting our approach. We define a semantic proximity allowing to separate the semantic interoperable information systems in segments that allows encircled efforts of query answering. Thus our approach consists to exploit intelligent agents for query reformulation and new technologies for semantic representation. Our query answering approach adapts to the changes of environment and solve the most various data conflicts in a dynamic way.

In the following, section 2 presents a synthesis of the various existing approaches. Two sections 3.1 and 3.2 describe the architecture of the mediation and the various types of agents. Then sections 3.3 and 3.4 explain the queries processing, the communication language between agents and the prototype implementation. 2. Related work

As The problem of query answering in distributed

systems has been discussed in traditional databases and the Semantic Web. There are two possible orientations to query answering: the integration guided by the sources (schema mediation), and the integration guided by the queries (context mediation) [21][5] [6][8][4][10][11].

23

mailto:[email protected]�

mailto:[email protected]�

A Scalable and Efficient Query Answering for a Context and Schema Mediation Benharzallah Saber, Kazar Okba

The schema mediation is a direct extension of the federate approach. Data conflicts statically solved. In the schema mediation; the mediator should associated with a knowledge set (mapping rules) for locating the data sources. The query answering follows a execution plan established by rules which determine where are the relevant data to treat a query. It requires a pre-knowledge on the systems participating in the cooperation. The mediator’s role is to divide, according to the global schema, the user query in several sub-queries supported by the sources and gathers finally the results. The global schema is generally specified by object, logic, XML or OWL interfaces [24][17][3][5][22]. In all these works, the objective is to build a global schema which integrates all the local schemas. When one operates in an evolutionary world where sources can evolve all the time, the elaboration of a global schema is a difficult task. It would be necessary to be able to reconstruct the integrated schema each time a new source is considered or each time an actual source makes a number of changes [4]. In general the response time of the queries of this approach is better than the context mediation which requires much time (it uses the semantic reconciliation). In this approach the transparency is assured. The degree of automation of the resolution of the data conflicts is weak, and the scalability is less respected, compared to the context mediation.

Very numerous works intervened to propose

automatic approaches of schemas/ontologies integration [30][31]. An approach was particularly investigated: the mapping of schemas. It led to the elaboration of several systems such as DIKE [7], COMA [13], CUPID [14]. One will find analyses and comparisons of such systems in [18]. Several approaches of integration of information based on ontologies were suggested. One will find a synthesis of it in [20][4]. On one hand, the drawback of these works is that they do not describe the integration process in a complete way; they always use assumptions like pre-existence mappings [23][33]... On the other hand they provide methods to calculate mappings between general or specific ontologies [30] and they do not indicate how to really exploit it for automatic integration or for the query reformulation [22][33].

In [21][3] Propose an extended schema mediation

named DILEMMA based on the queries static resolution. The mediation is ensured by a couple mediator/wrapper and a knowledge base associated with each system that takes part in the cooperation. The mediator comprises a queries processor and a facilitator. This approach provide a better transparency and makes it possible to solve the semantic values

conflicts, but in a manner a priori. The automation degree is better compared to the schema mediation which remains always indeed to the recourse of an expert of the domain. It has a low capacity to treat evolutionarily and the scalability.

First tendency can be illustrated using work of

Kashyap and Seth (1994) for the context mediation approach. Sensitive to the semantic heterogeneity problem which can make difficult query processing, they introduced the concept of context. The role of the mediator in this approach, is to identify, to locate, transform and integrate the relevant data according to semantics associated with a query [21][3]. The data conflicts resolution is dynamic and does not require the definition of a mediation schema. The user's queries are generally formulated in terms of ontologies. The data are integrated dynamically according to the semantic information contained in the description of the contexts. This approach allows the best taken into account of the evolution of the local sources and the automation degree is better compared to the schema mediation. Two categories of context mediation are defined: - the single domain approach SIMS [9], COIN [10] works on a single domain where all the contexts are defined by using a universal of consensual speech. The scalability and evolutionarily are respected but remains limited by the unicity of the domain. - Multi-domains approaches Infosleuth [11], Observer [12] they use various means to represent and bind heterogeneous semantic domain: ontologies, hierarchy of ontologies and method of statistical analysis. In the context mediation approach the data conflicts are dynamically solved during the execution of the queries, allowing the best evolution of the local sources and the automation degree is better compared to the schema mediation, this with the detriment of response time of the queries (it uses the semantic reconciliation). Concerning the semantic conflicts, the majority of the projects solve only the taxonomic conflicts Coin [10]. The resolution of the values conflicts either is guided by the user Infosleuth [11], or unsolved in the majority of the cases Observer [12] [28].

The agent paradigm gives a new sight for the development of the of complex, heterogeneous, distributed and/or autonomous systems nature [15][34][35][38]. Several works of semantic interoperability use the agent paradigm [16]-[11][25][29] [32].

Infosleuth project [11] is to implement a set of cooperative agents which discover, integrate and present information according to the needs for a user or an application for which they produce a simple and coherent interface. The Infosleuth architecture project consists of a set of agents collaborative communicate

24


by agent communication language KQML. The users express their queries on a specific ontology. KIF and SQL are used to represent the queries. The queries are to dispatch with the agents specialized (agent broker, ontological, planner...) for the data retrieval on distributed sources and for integration. The resolution of many semantic conflicts remains guided by the user [3]. They use specialized agents which can see as threads are far from the definition of the cognitive agent like defined in the distributed artificial intelligence.

In [25] propose a multi-agent system to achieve semantic interoperability and to resolve semantic conflicts relatively to evolutive domain ontologies. In this approach the query answering and the validation of the mappings are completely related to the users. [29] Propose an agent based intelligent meta-search and recommendation system for products through consideration of multiple attributes by using ontology mapping and Web services. This framework is intended for a electronic commerce field.

3. Approach description

The objectives of our solution are the realization of

a semantic mediation having the following characteristics:

1) To allow a system which provides its context of application, to find the information systems and information shared by those systems which are useful to this application and to use them in a transparent way.

2) To ensure the advantages of the context and schema mediation, and to avoid their disadvantages. Our approach focuses on the dynamic change of the multi-agents mediation system, of the context mediation to the schema mediation during the operation of the mediation system, that is done by the use of the intelligent agents, in order to ensure the open aspect of the context mediation (high automation degree of the query resolution process...) and the robustness of the schema mediation (reliability of the answers by the use of the mapping rules, the formulation of the queries in schema terms ...).

3) To solve the majority of the semantic conflicts by using new technologies for semantic representation, and by respecting a high automation degree of the query answering. Our query answering approach balances the trade-off between performance and completeness by introducing segmentation based semantic interoperable information systems. We define the notions of semantic proximity that allows

encircled efforts of query answering. Moreover, the query answering algorithm reformulates the query by using the schema mediation method for the systems discovered and the context mediation for the others. This fact, our query answering approach adapts to the changes of environment.

3.1 A generic architecture based agent of semantic mediation

The cooperation suggested in our solution based on

a preliminary construction of information having to be integrated and on the mixed query resolution static and dynamics. An information system can play in turn the role of supplier or consumer [Figure1] of information. Our architecture consists of two types of agents: intelligent agents and routings agents (described below)

The integration phase, of a new information system in our proposed mediation system, begins with creation of an intelligent agent and continues with the fastening of this last to a routing agent which is nearest semantically (algorithm 4).

The intelligent agent applies the Contract Net protocol. The intelligent agent sends an invitation to tender describing its domain. The routing agents receiving the call provide their ability (semantic proximity rate, algorithm 3). As soon as, the intelligent agent receives answers of the all routings agents, it evaluates these rates, and makes its choice on a routing agent which is the nearest semantically. The chosen routing agent adds the intelligent agent to its net contacts.

Routing Agent Routing Agent

Intelligent Agent

Intelligent Agent

User/Application

Supplier IS

Multi agent cooperation

Intelligent Agent

Supplier IS

Intelligent Agent

Supplier IS

Consumer IS

Intelligent Agent

User/Application

Consumer IS

Intelligent Agent

User/Application

Consumer IS

Figure 1. General architecture

Our approach does not use a global schema or some predefined mappings. The users interrogate the

25


consuming system (the queries are formulated in the consuming schema terms).

At the beginning the intelligent agent consuming

(IAC) applies the protocol of the dynamic query resolution (context mediation) because it does not have information on the suppliers systems, during the dynamic evaluation of the query, the intelligent agent suppliers (IAS) updated their histories and add information (mapping between terms of query ontology and IAS) to facilitate their dynamic integration with the IAC.

During the operation of the mediation system, the

IAC applies the protocol to discover suppliers which the nearest semantically with its domain, and to integrate them dynamically in order to use them in the schema mediation. For that, it cooperates with the routing agent. The IAC updates its knowledge base by mapping rules.

During the operation of the system, the IAC

discovers certain suppliers and adapts with the environment. So, to treat a query, it applies the two protocols: static query resolution for the systems discovered and the dynamic query resolution for the others systems.

3.2 Description of the various types of agents We first introduce some concepts and definitions

necessary to explain our agents and query answering. 3.2.1 Definitions

Definition 1: Schema a schema describes the structure of a database.

Our approach use the OWL[19] (Web Ontology Language) as common data model and we base on the work[17] which uses OWL like ontologies representation model. The OWL enriches the RDF Schemas model by defining a rich vocabulary; for the description of complex ontologies. So it is more expressive than RDF and RDFS which have some an insufficiency of expressivity because they depend only on the definition of the relations between objects by assertions. OWL brings also a better integration, an evolution, a division and an easier inference of ontologies [19].

For example, the relational schema is transformed into: The relations are represented by OWL classes, the relations between classes are represented in OWL by owl:ObjectProperty and owl:DatatypeProperty.

OWL properties can represent the various attributes and constraints contained in the schema. Properties OWL represent the attributes by Datatype. If the attribute is a primary key then the functional property will be added. One use of owl:ObjectProperty is to represent the attributes which are under a constraint of foreign key.

Definition 2 Ontologies

In this paper, an ontology described the concepts of a domain.

Given a set iC of classes and a set

iP of

properties in ontologyiO . iO is a set of axioms of the

following forms: • 1iC ⊆ 2iC , 1iP ⊆ 2iP , or • 1iC ≡ 2iC , 1iP ≡ 2iP , Ontologies are described by subsumption and

equivalence relations for classes and properties. We distinguish in our approach three types of

ontologies: - An ontology built starting from a schema, -An ontology describe the name of an application,

(definition 11) - An ontology built starting from a query (semantic

enrichment of a query, definition 13)

Definition 3: Schema-ontology mapping Given a schema S and its ontology O. a schema-

ontology mapping is the function: S-O: S → O x → e We can see the elements of a schema S as

individuals of the classes defined in ontology O.

Definition 4: Context A context describes the assumptions and explicit

information of definition and the use of a data [3]. In our approach the context is defined by (S, O, S-O) such as: S a schema, O an ontology and S-O is a schema-ontology mapping.

Definition 5: Query language

We adopted the language defined in [2] as query language in our architecture Let V is a set of variables disjoint from L. A query

iQ in ontology iO is of the form ,i

PiC QQ ∧ where

• iCQ is a conjunction of )(xC i where CC i ∈ and

VLx ∪∈ • i

PQ is a conjunction of ),( yxP i where PP i ∈ and VLyx ∪∈,

26


Definition 6 The semantic distance (ds) between two elements

e1 and e'1, of two ontologies, of two different agents, is the rate of resemblance between these two elements. The calculation of this distance based on a taxonomic analysis of the terms (list of the synonyms and antonyms) which characterize the two elements and on also the vicinity of the two elements.

Definition 7

The comparison of two ontologies, from two different agents, is only the finding correspondence between their entities which are semantically similar. The comparison is defined by the Comp function as follows: Comp : 'OO → such as Comp(e1) = e'1 if ds(e1;e'1) > tr where O and O' are two ontologies to be compared, tr indicates a minimal threshold of similarity belonging to the interval [0,1], e1∈O and e'1∈O'. e1 and e'1 represent the entities of two ontologies. The tr indicates the minimum level in witch two entities are similar.

Definition 8

- Given a two agents A, B. - Given the schema aS and ontology aO of agent A.

Thus, the ontology bO of the agent B. - Given the function Comp: ba OO → it is the

comparison between two ontologies aO and bO of A and B respectively.

- Given abCO is the set of the elements e∈ aO , such as Comp(e) = e' and ds(e; e') > tr with e’∈ bO .

- Given a sub-schema aSs is the set of elements

x∈ aS such as S-O(x) =e with e∈ abCO .

The adaptation of sub schema aSs of aS (of the agent A) on ontology bO of the agent B is the function:

Adapt: →aSs bO →x e' With: x∈ aSs , e'∈ bO where exist e∈ abCO such

as Comp(e) = e' and ds(e; e') > tr, S-O(x) =e .

Definition 9: Mapping rules A schema mapping is a triple (S1, S2, M)[2], where:

S1 is the source schema; S2 is the target schema; M the mapping between S1 and S2, i.e. a set of assertions

Ts qq , where sq and Tq are conjunctive

queries over S1 and S2, respectively, with the same set of distinguished variables x, and { }≡⊇⊆∈ , , .

Definition 10: Query Reformulation Let

iQ be a query in schema iS and

jQ be a query in schema

jS described by classes and properties in the mapping Mij .

• jQ is an equivalent reformulation of

iQ if

ij QQ ⊆ and ji QQ ⊆ . noted

ij QQ ≡ .

• jQ is a minimally-containing reformulation of iQ

if ji QQ ⊆ and there is no other query '

jQ such that 'ji QQ ⊆ and

jj QQ ⊆' .

• jQ is a maximally-contained reformulation of

iQ if

ij QQ ⊆ and there is no other query 'jQ such that

'jj QQ ⊆ and

ij QQ ⊆' . To find the approximate query reformulation we use

the algorithm defined in [2].

Definition 11: Application domain description Is the set of all the terms, organized in the form of

an OWL ontology, describing the name of the application domain.

Definition 12: Semantic proximity

Is a measurement between two OWL ontologies of two application domains. Given two application domains 1d , 2d and their ontologies 1dO and 2dO respectively. The semantic proximity is a function SemProx: ]1,0[21 →× dd OO . This function calculates the semantic proximity between the two names 1d and

2d , by using two ontologies 1dO and 2dO . The algorithm used to calculate this proximity is ASCO2 [37][36].

Definition 13: Semantic enrichment of a query

Given the context C represent by (S, O, S-O). Given Q= ,i

PiC QQ ∧ a query formulated in term of the schema

S. The semantic enrichment of this query, by using

ontology O, is defined by the following rules: 1) Find by using the function S-O, the equivalent

classes of )(xCi and ),( yxP i of the query iCQ and i

PQ respectively in the ontology O. They respectively are noted )(xOC i and

),( yxOPi

. 2) Find by using the subsumption relation, the

ancestors classes of each class of )(xOC i and

27


),( yxOP i . They respectively are noted

)(xpOCi and ),( yxpOP i . 3) Find by using the subsumption relation, the sub

classes of each class of )(xOC i and

),( yxOP i . They respectively are noted

)(xcOCi and ),( yxcOPi . 4) Find by using the equivalent relation, the

equivalent classes of each class of )(xOC i and ),( yxOP i . They respectively are noted

)(xeOCi and ),( yxeOPi . A query Q enriched semantically is composed of

( ,iP

iC QQ ∧ , )(xOC i ,{ )(xeOCi },{ )(xpOCi },{ )(xcOCi },

),( yxOP i ,{ ),( yxeOPi },{ ),( yxpOP i },{ ),( yxcOPi }). We can see it as an query ontology.

Definition 14: Semantic evaluation of a query

The semantic evaluation of a query enriched semantically QO = ( ,i

PiC QQ ∧ , )(xOC i ,{ )(xeOCi },{ )(xpOCi },{ )(xcOCi },

),( yxOP i ,{ ),( yxeOPi },{ ),( yxpOP i },{ ),( yxcOPi }) is defined by algorithm 1(figure 2):

Figure 2. Algorithm 1. Semantic evaluation of a

query 3.2.2 Description of the agents

Intelligent agents

They are mono-domain agents. They acquire

information coming from other agents; they adapt gradually and grow rich starting from their internal knowledge bases and evolution coming from the system. The intelligent agent roles are multiple:

1) Execution of the queries of the users or applications.

2) The comparison of ontologies and the automatic generation of the mapping rules of the schemas between an IAC and another intelligent agent supplier (IAS) (figure 3).

Figure 3. Algorithm 2. Generation mapping rules

3) It enriches the query semantically by using the

ontology and the links schema-ontology (definition 13).

4) It translates the query Q coming from other IACs into query Q' expressed in the proper language of handling of the local base of the supplier.

5) Filtering of the results.

Routing agents

They are multi-domain agents, gathering the nearest semantically domain. The roles of a routing agent are:

1) To gather the nearest semantically intelligent agents in a net contacts to be used as suppliers (figure 4 and 5).

Algorithm 2. Generation mapping rules Require: a schema, ontology and schema-ontology mapping of IAC and IAS 1: Calculation the function Comp: IASIAC OO →

2: Calculation the set IACIASCO

3: Calculation the sub-schema IACSs , IASSs

4 : Calculation the function Adapt : IACSs IASO 5 : /* Mappings generation */ 6: For each x ∈ IACSs do

7: Search y ∈ IASSs such as:

8: If Adapt(x)=e1 and IASOS − (y)=e2 and e1, e2 ∈

9: IASO and if e1⊆ e2 then generate the rule x⊆ y

10: else if e1⊇ e2 then generate the rule x ⊇ y else 11: if e1≡ e2 then generate the rule x≡ y 12: EndFor 13: End

Algorithm 1 Require: query ontology QO of IAC and ontology

IASO of IAS

1: Calculation the function Comp: IASQ OO →

2: Calculation the set QIASCO

3: Calculation the sub-schema QSs

4 : Calculation the function Adapt : QSs IASO

5: Evaluate the semantic query Adapt( QSs ),

6: End

28


Figure 4. Algorithm 3. Semantic proximity rate

Figure 5. Agorithme 4. Nearest semantically routing agent

2) To ensure the dynamic query resolution, and to communicate with other routing agents for to execute reformulated queries.

3) To record / eliminate dynamically the agents which take part in the cooperation.

4) To search intelligent agents which contain information to which the domain is the nearest to the domain of an intelligent agent (consuming).

3.3 Queries processing

The query processing is divided into several steps, and during this process, the multi-agents system uses a set protocol. The principal steps are (figure 6): 3.3.1 Static query resolution

The static resolution is made for the systems which are already discovered.

Step 1: query validation the IAC checks the validity of the query. i.e. whether is written in schema mediation terms or not,

Step2: query reformulation: the query is divided into a recombining query of the results and sub queries intended for the IAS which contain data necessary to the execution of the query. The decomposition of the query is done by the use of the mapping rules. The IAC

Step3: recombining of the results: the IAC execute the recombining query of the results.

applies the cooperation protocol of static query resolution.

3.3.2 Dynamic query resolution

The dynamic resolution makes it possible to take into account the appearance of new IASs. The principal steps are:

Step 1: Semantic enrichment of a query. The IAC enriches the query semantically by using the ontology and the links schema-ontology which are in its knowledge base (definition 13).

Step 2: Transmission of the semantically enriched query. The IAC applies the cooperation protocol of dynamic query resolution. So it transmits the semantically enriched query to the routing agent which is nearest semantically. This latter sends it to all IASs of its net contacts.

Step 3: Semantic evaluation of the semantically enriched query. Each IAS answers according to its capacity to treat the query:

1) To compare elements of the query with its ontology. The elements of the query and its ontology are compared by using a semantic distance. The identified elements as equivalent are retained.

2) The query is rewritten in terms of the equivalent elements of its ontology (then interpreted on its schema) to take of them into account the semantic conflicts of values (each intelligent agent has library of functions for the conversion of the types) (definition 14).

3) The answer is sent latter to the routing agent, indicating the manner of treating the query, for this letter can build recombining queries of the results.

If no IAS answers, the routing agent sends the query to the other routings agents of other domains and if there are answers the routing agent updates its net contacts.

Stage 4: Results recombining: the routing agent recomposes the results obtained of IASs. Then it sends

Agorithme 4. Nearest semantically routing agent Require: the intelligent agent receives answers of the all routings agents; L is the list of these agents contains its semantic proximity rate. 0: NSRA0; 1: while L<> ∅ do 2: extract an element e from the list L 3: if e.rate > NSRA then 4: NSRA e.rate 5: Save the routing agent who has this rate 6: Endif 7: endWhile 8: return the routing agent saving

Algorithm 3. Semantic proximity rate Require: a ontology describe the domain name of intelligent agent and ontologies of the others, mapping of IAC and IAS 0: Sim0 ; 1: for each Θ∈iO do /* i varies from 1 to N */

2 : SimSim+SimProx( newIAO , iO ) 3 : EndFor 4 : Sim Sim/N 5 : semantic proximity rate is Sim 6: end.

29


the final result to the IAC, this latter recomposes the results of static and dynamic query resolution.

Figure 6. Algorithm 5. Query answering

3.4 Prototype implementation on the multi-agents JADE framework

In our solution we adopted the ACL of the FIPA like a communication language between agents. FIPA-ACL [26] has on KQML the advantage of a more rigorous semantics as well as a significant effort of standardization. Our agents use a cooperation protocols. The cooperation protocols were established on the Jade framework (Java Agent Development Framework) which provides API to develop a multi-agents system. This framework was selected because it proposes a library of interaction protocols. The addition of new information sources implies the creation of new agents on the Jade framework. The addition of the sources in great quantity concerns the property of scalability. The importance of the number of agents requires a more significant number of communications, which affects the request of the transport system of the Jade messages. The scalability and the performances of the transport system of Jade message were treated in [27][28]. The results obtained confirm the fact that Jade milked well the scalability according to several scenarios (will intra or inter framework). 4. Conclusion and prospects

Establishing semantic interoperability among heterogeneous and distributed information systems has been, and still is, a critical issue attracting significant attention from research and practical reality.

In this paper we presented a generic, scalable and efficient query answering for semantic interoperable information systems. We proposed a generic multi agent architecture supporting our approach.

The main advantage of our query answering is its robustness with regard to the evolution of systems,

adaptation to the changes of environment and solves the most various data conflicts in a dynamic way.

Our approach divides the interoperables information systems into sub interoprables information systems (or segments). These sub systems share a application domains nearest semantically. These sub systems are built gradually at the step of recording of the new intelligent agents. The treatment of a query of a IAC is done on the segment that the IAC belongs, except in case where there does not exist results (section 3.3.2). The segmentation allows the facility and the speed of the treatment of the queries, on avoiding to send and to treat useless messages. It allows also the scalability of the system.

Our approach has double effects, it treats at the beginning the queries of users by using the context mediation, and at the same time it tries to discover information systems and to integrate them dynamically.

This fact our approach causes the multi agents mediation system to use schema mediation, to avoid the semantic reconciliations (used in the context mediation which take much time). And in order to take into account the open and dynamic aspect of the environments and also to ensure the scalability, our approach uses the context mediation for the systems not discovered.

The developed prototype shows us the functionality of architecture suggested. As prospect we try to slacken our data mediation towards service mediation in general and to use intelligent methods to reduce ontologies to be compared for not to influence the scalability of architecture suggested. 5. References [1] Journal article : E. Sciore, M. Siegel, A. Resenthal,

Using Semantic Values to Facilitate Interoperability Among Heterogeneous Information Systems, ACM Transaction on Database Systems, Vol. 19, N.2, pp. 254-290, Jun. 1994

[2] Jun-ichi Akahani, Kaoru Hiramatsu, and Tetsuji Satoh, Approximate Query Reformulation for Ontology Integration, Semantic Integration Workshop (SI-2003) Second International Semantic Web Conference October 20, 2003 Sanibel Island, Florida, USA

[3] Thesis: Fabrice Jouanot, « DILEMMA : vers une coopération de systèmes d’informations basée sur la médiation sémantique et la fusion d’objets », université de bourgougne, novembre 2001.

[4] Michel Schneider, Lotfi Bejaoui, Guillaume Bertin, A Semantic Matching Approach for Mediating heterogeneous Sources, metadata and semantic, springer US, p 537-548, ISBN 978-0-387-77744-3, 2009

[5] Article: Susanne Busse, Ralf-Detlef Kutsche, Ulf Leser, Herbert Weber, “Federated Information Systems: Concepts, Terminology and Architectures”, Technical Report Nr.99-9 Berlin University,1999

Algorithm 5. Query answering Given L the list of discovered agents and their mappings If QueryValidation() then 1:if L <> empty then

- QueryReformulation() - StaticRecombiningResults()

2: Dynamic query resolution - SemanticEnrichmentQuery() - TransmissionSemanticallyEnrichedQuery() - SemanticEvaluation() - DynamicRecombiningResults()

30


[6] Farshad Hakimpour, thesis Using Ontologies to Resolve Semantic Heterogeneity for Integrating Spatial Database Schemata, Universität Zürich Zürich, 2003

[7] Palopoli, L., Terracina, G., Ursino, D.: DIKE: a system supporting the semi-automatic construction of cooperative information systems from heterogeneous databases. Softw., Pract. Exper. 33(9), pp. 847-884 (2003).

[8] Eric LECLERCQ, Djamal BENSLIMANE, Kokou Yétongnon, "ISIS: A Semantic Mediation Model and an Agent Based Architecture for GIS Interoperability", in Proceedings of the International Database Engineering and Applications Symposium (IDEAS'99), Montreal, CANADA, August 2-4, 1999. Published by IEEE Computer Society Press, ISBN 0-7695-0265-2, p. 81--92

[9] ARENS Y. et al , Query processing in the SIMS Information Mediator. In TATE A. Ed. Advanced planning technology, Menlo park, CA, AAAI Press, 1996.

[10] Bressan S, et al., “Context Knowledge Representation and Reasoning in the Context Interchange System”, Applied Intelligence, vol. 13, n°2, September 2000, p. 165-180.

[11] Article: R. J. Bayardo Jr.et al, “InfoSleuth: Agent-Based Semantic Integration of Information in Open and Dynamic Environments”, Microelectronics and Computer Technology Corporation,1997

[12] Article: Mena E., Illarramendi A., Kashyap V., Sheth A., “OBSERVER : An approach for Query Processing in Global Information Systems based on Interoperation across Pre-existing Ontologies”, International journal on Distributed And Parallel Databases DAPD, vol.8, n°2, April 2000, p. 223-272.

[13] Do, H. H., Rahm, E.: COMA - A System for Flexible Combination of Schema Matching Approaches. VLDB 2002, pp. 610-621 (2002).

[14] Madhavan, J., Bernstein, P.A., Rahm, R.: Generic Schema Matching with Cupid. VLDB 2001, pp. 49-58 (2001).

[15] R. GIRARDI, "An analysis of the contributions of the agent paradigm for the development of complex systems", In (SCI 2001) and (ISAS 2001), Orlando, Florida. 2001.

[16] Purvis M, Cranefield S., Bush G., Carter D., McKinlay B., Nowostawski M. Ward R., The NZDIS Project: an Agent-Based Distributed Information Systems Architecture, Proceedings of the Hawai’i International Conference On System Sciences, January 4- 7, 2000, Maui, Hawaii.

[17] Seksun Suwanmanee et al, OWL Based Approach for Semantic Interoperability, Proceedings of the 19th International Conference on Advanced Information Networking and Applications (AINA 05)

[18] Mohsenzadeh, M., Shams, F., Teshnehlab, M.: Comparison of Schema Matching Systems. WEC (2), pp 141-147 (2005).

[19] Deborah L. McGuinness and Frank van Harmelen. « Owl web ontology language ». Technical report, 2004.

[20] Wache, H., Vogele, T., Visser, U., Stuckenschmidt, H., Schuster, G., Neumann, H., Hubner, S.:Ontology-based integration of information - a survey of existing approaches. In Stuckenschmidt, H., ed., IJCAI-01 Workshop: Ontologies and Information Sharing, pp 108-117 (2001).

[21] Fabrice Jouanot, Nadine Cullot, and Kokou Y´etongnon, Context Comparison for Object Fusion CAiSE 2003, LNCS 2681, pp. 536−551, 2003.

[22] Rubén Tous, Thesis: Data Integration with XML and Semantic web Technologies. Universitat Pompeu Fabra, Barcelona, June 2006

[23] François Goasdoué and Marie-Christine Rousset, Querying Distributed Data through Distributed Ontologies: ASimple but Scalable Approach, IEEE INTELLIGENT SYSTEMS,p60-65, 2003

[24] Nang Tom Hawm Saw, OWL-Based Approach for Semantic teroperating and Accessing Heterogeneous XML Sources, 0-7803-9521-2/06/$20.00 §2006 IEEE.

[25] Guilaine Talens, Danielle Boulanger, and Magali Séguran, Domain Ontologies Evolutions to Solve Semantic Conflicts M. Collard (Ed.): ODBIS 2005/2006, LNCS 4623, pp. 51–67, 2007., Springer-Verlag Berlin Heidelberg 2007

[26] Online document: FIPA Communicative Act Library Specification, 2000

[27] E. Cortese, F. Quarta, G. Vitaglione. Scalability and Performance of Jade Message Transport System. Presented at AAMAS Workshop on AgentCities, Bologna, 16 th July, 2002.

[28] Thesis: Magali Séguran, Résolution des conflits sémantiques dans les systèmes d'information coopératifs : proposition d'un modèle d'interaction entre agents, Université Jean Moulin, Lyon3, 2003

[29] Wooju Kim & Dae Woo Choi & Sangun Park, Agent based intelligent search framework for product information using ontology mapping, J Intell Inf Syst (2008) 30:227–247 DOI 10.1007/s10844-006-0026-8, Springer Science + Business Media, LLC 2008

[30] Fleur Mougin, Anita Burgun, Olivier Bodenreider, Julie Chabalier, Olivier Loréal4, and Pierre Le Beux1, Automatic Methods for Integrating Biomedical Data Sources in a Mediator-Based System, DILS 2008, LNBI 5109, pp. 61–76, 2008. Springer-Verlag Berlin Heidelberg 2008

[31] H. L. Hansen, K. Ingvaldsen , Dynamic Schema Integration The ACRoS Prototype THESIS, The Faculty of Social Sciences, Department of Information Science and Media Studies, may 2005

[32] Avigdor Gal and Aviv Segev, Agent Oriented Data Integration, J. Akoka et al. (Eds.): ER Workshops 2005, LNCS 3770, pp. 98–108, 2005.Springer-Verlag Berlin Heidelberg 2005

[33] Peter haase,dissertation, semantic technologies for distributed information systems, universitatsverlag karlsruhe 2007, ISBN 978-3-86644-100-2.

[34] Ning Zhong, Jiming Liu, Setsuo Ohsuga, Jeffrey Bradshaw, Intelligent Agent Research and Development 2nd Asia-Pacific Conference on IAT , ISBN 981-02-4706-0 World Scientific Publishing Co. Pte. Ltd., 2001

[35] J. G. Carbonell and J. Siekmann, Lecture Notes in Artificial Intelligence, Agent-Oriented Information Systems III 7th International Bi-ConferenceWorkshop, AOIS 2005 Utrecht, ISBN-10 3-540-48291-1 Springer Berlin Heidelberg NewYork, 2006

[36] Bach, T.L., Dieng-Kuntz, R. et Gandon, F. On Ontology Matching Problems (for building a corporate Semantic Web in a multi-communities organization). Proceedings of ICEIS 2004. Porto, Portugal. 14-17 April 2004.

[37] Bach, T.L. et Dieng-Kuntz, R. Measuring Similarity of Elements in OWL Ontologies. AAAI’2005 workshop on

31


Contexts and Ontologies: Theory, Practice and Applications. Pittsburgh, Pennsylvania, USA, July

[38] LI Ai-Ping, Jia Yan and WU Quan-Yuan, "A Study on Organizational Knowledge in Multi-Agent System", JCIT:Journal of Convergence Information Technology, Vol.2 No.1, pp.61-65, 2007

32

a scalable and efficient query answering for a context and ... · pdf filea scalable and...

Documents