=Paper=
{{Paper
|id=Vol-160/paper-9
|storemode=property
|title=An Architecture for Heterogenous Federated Mediators
|pdfUrl=https://ceur-ws.org/Vol-160/paper7.pdf
|volume=Vol-160
|dblpUrl=https://dblp.org/rec/conf/caise/ChengB05
}}
==An Architecture for Heterogenous Federated Mediators==
https://ceur-ws.org/Vol-160/paper7.pdf
An Architecture for Heterogeneous Federated
Mediators
Dong Cheng and Nacer Boudjlida
LORIA-University Henri Poincaré Nancy 1
BP 239, 54506 Vandoeuvre Lès Nancy CEDEX(F)
Email: cheng,nacer@loria.fr
Abstract. This paper describes a heterogeneous federated mediators
architecture to support querying and cooperation activities. In a lot of situations,
the data and program were recorded by different persons, at different times, in
different languages, on different fields. All the difference makes the
information systems’ heterogeneity. The mediation services are expected to
play an important role in helping automated processes to access heterogeneous
information. They are not Yes/No answers from the mediator, when no single
object meets the search criteria, they may be cooperative answers to make a
composite answer. In this paper we introduce a mediator-based logical
architecture and its implementation.
1. Introduction
The propagation of the network has led to increasing need for interoperability
between heterogeneous systems and services. These latest developments in data
network domain are new challenges for interoperability technology, as attested by the
tremendous work about Semantic Web [1]. However, in this work we in-tend to
define an approach where heterogeneous systems may work together in a context of
semantic queries. In this condition, we must match the information that was created
by different programmers, at different times, in different languages, and all these
things may be described in different knowledge representation technologies.
Examples of applications of our work may be searching for Web Services that offer
given functions, searching for a component in the context of component-based design
and component-based programming, searching for a business partner with a given
expertise, looking for an employee whose records and expertise satisfy a given
position profile...
In this paper, we highlight a Composite Answer approach in heterogeneous
environment where systems maybe record the information in different knowledge
representation technologies. A significant originality of our approach resides in the
type of answers we aim at providing. Indeed, when there is no unique entity to satisfy
the search criteria, the systems attempt to determine a composite answer that satisfies
the criteria. Possibly heterogeneous systems may cooperate in the query evaluation
process and in the answer composition process. These systems are viewed as a
federation.
We are exploring alternative architecture, component and mechanisms to
transparently federate such types of systems or services. This paper reports on a first
experimentation focusing on the architecture itself. The presentation is structured as
follows. In section 2 we briefly review the conceptual federated mediator-based
architecture, and the concepts of composite answers in this architecture. In section 3,
we describe an implementation and the physical architecture. Conclusions and
remarks are in section 4.
2. A Review of Mediation Architecture
Exporter
LTell System 1 LComplement System 2 LComplement
y
er
L Qu Knowledge Knowledge
Base LAnswer Base LAnswer
LAnswer
Importer
Fig. 1 The Mediator-based Architecture
In our work a mediator-based architecture has been adopted and is described in [7].
It is a dynamic discovery of services or capabilities an “entity” offers. It is very
similar to the notion of discovery agency in the Web services architecture. In this
architecture, an “entity”, called exporter, publishes (tells) its capabilities at one or
more mediators sites (see figure 1). Entities, called importers, send requests (queries)
to the mediator asking for exporters fitted with a given set of capabilities.
Several approaches, which are mediator-based, i.e. they are distributed and
intelligent production systems, have been proposed over the last decade. A single
Mediator is designed to offer an adequate level of decision-making integration, it
takes into account the effort needed for the integration of heterogeneous computer
systems [8]. The Conflict Resolution Environment for Autonomous Mediation
(CREAM) system has been implemented, it provides various user groups with an
integrated and collaborative facility to achieve semantic interoperability among
participating heterogeneous information sources [9]. The KRAFT (Knowledge Reuse
And Fusion/Transformation) architecture provides a generic infrastructure for
knowledge management applications. It supports virtual organization using mediator
agents [10]. When it was applied to business-to-business electronic commerce, the
mediators allow partners to exchange rich business information. As online travel
agent application, a nature of competition in electronic markets, when it gets a
customer’s travel requirements to reservation system, envision this hypothetical
system designed for a group of airlines all of whom serve national routes. Consider
what might occur if that system was extended to include international airlines and
foreign routes. These travel requirements could be easily satisfied by one travel agent
in federated mediators.
In federated mediators, a cooperation environment, we can see the query as being
addressed to “the union” of the federated mediators’ knowledge bases. Concretely,
this union is explored from “near to near” within the federation, that means from a
mediator to another. Satisfying the query falls into different cases [7]:
– Case1: There exist exporters that exactly satisfy the query;
– Case2: There exist exporters that fully satisfy the query, but their
capabilities are wider than those requested;
– Case3: No single exporter fully satisfies the query, but when “combining”
or composing capabilities from different exporters, one can fully satisfy the
query;
– Case4: Neither a single exporter nor multiple exporters satisfy the query, but
there exist some exporters that partly satisfy the query;
– Case5: No single exporter nor several exporters fully or partly satisfy the
query.
One notices, that cases 3 and 4 are cases where a composite answer is re-turned.
In the case 4, we have to determine “what is missing?” Complement Concept, to the
individuals to satisfy Q, which means to determine what part of the query is not
satisfied by the found individuals. Furthermore, in these cases 2, 3 and 4, we need to
notice “what is superfluous?”
An approach to find out “the best” answer was proposed using Description Logics
[11]. The determination of the complement is based on the subsumption relationship
using a Normalize-Compare process. The implementation of this process uses an
array of Boolean (called “Table Of Test” further) to record the results of the
subsumption relationship evaluation. In the figure 2, C1, C2, C3,… , Cn denote the
query concept under its normal form and D1, D2, D3,… , Dm denotes the concepts
“known from” the mediators, i.e. every Dj has to be viewed under its normal form
Dj1,Dj2, …,Djn.
Then TableOfTest[Dj,Ci] = true means that . When the value re-turned by
the function Subsumes(C,Dj) is “false” (i.e. the concept Dj does not fully satisfy the
concept C.), therefore we need to determine a possible complement of Dj relatively to
C. Using the Table Of Test it is easy to get the complement of the concept Dj
relatively to the concept C:
Comp(C,Dj) = . That means that the
complement is given by the conjunction of all the atomic concepts for which the
corresponding values in the “Table Of Test” are “false”.
The composition of the truth values determines the cases of satisfaction. Consider
a table ORoD[1..n] as ORoD[i] = . ORoD[i] = true means that the
concept Ci is satisfied by at least a Dk. If the conjunction of the values of ORoD,
Fig. 2 The “Table Of Test”
noted ANDoS, is true (i.e. ), it means that all the Cis are
satisfied and therefore the query. When ANDoS is false, the logical disjunction of the
values of ORoD, noted ORoS, enables to determine a possible partial satisfaction: if
, it means that there exist some Ck that are satisfied. If
both ORoS and ANDoS are false then no atomic concept Djk (j ∈ 1..m) satisfies a Ci.
For example, a formulaic query of travel requirement is in Description Logics.
It exists a mediator system 1, it well knows all routes in France and some inter-
national airlines. So it maybe finds a route as:
System 1 does not know any route in China, this travel requirement can’t be fully
satisfied. But we can find another mediator system 2, it well knows all routes in China.
So it easy to find a route as:
We put the D1 and D2 into a Table Of Test. We get ANDoS=True, it means that
this travel requirement can be fully satisfied by a composite answer of D1 in system 1
and D2 in system 2.
3. Implementation and Architecture
Mediator 3
Local Reasoning
Repository Processor
Exporter
lexical Syntax
ontological Translator
Importer dictionary
Network
Mediator 1
Reasoning Local
Processor Repository Web Server Mediator 2
Local Reasoning
Repository Processor
Syntax Translator lexical lexical Syntax
ontological Translator
ontological dictionary
dictionary
Fig. 3 The Layered Mediator Architecture
An experimental platform has been developed in Java where some services, like
testing the subsomption relationship, determining the complement concept, and
making a composite answer, have been implemented. As the drawing in figure 1, the
federated mediators work in distributed and heterogeneous environment. The
federated mediators architecture conforms to the Reference Model for Open
Distributed Processing(RM-ODP). As shown in figure 3, the mediator architecture
consists of three main parts: a local repository, a reasoning processor, a syntax
translation.
In this implementation, we accept the Description Logical language AL- to reason
and represent knowledge. It is used to represent the entity's “capability” in the local
repository, and Java objects of AL- are used in all algorithms implementation. All
services were implemented in a Description Logic reasoning processor, using the
approaches and the algorithms that were outlined in the previous section.
3.1 The export sketch and inter-action
The exporter sends its capability description in AL- to the Reasoning Processor,
as showed in figure 4. In this situation, exporter accesses the Mediator server through
a Web Server CGI program, such as “classical” Web client/server architecture. The
servers accept the AL- concepts written in DAML+OIL, an ontology language in XML
[16]. The AL- concepts are encoded in DAML+OIL, then this DAML+OIL text is
transmitted to the Web Server's URL. The CGI} program may manage the
export/import task. It uses the Syntax Translator to translate the DAML+OIL text into
the AL- concept object. Then it puts the concept object into the Reasoning Processor.
The Reasoning Processor uses the Classification approach to add this concept into the
knowledge hierarchy (TBox and ABox) in the local repository.
Exporter Web CGI Reasoning Processor Syntax Translator TBox
http://mediator/export
daml2al()
Message1
export()
classify()
Fig. 4 The UML Sequence of Export Process
3.2 The import sketch and inter-action
The Reasoning Processor processes the query input from Importer like the
process of Exporter's capability concept, as showed in figure 5. Then this normalized
query concept is put into a TBox object. The algorithms in last section are
implemented in the TBox class. The services' result outputs are also encoded in
DAML+OIL. XML is also used to exchange information and concepts between the
mediators when mediators' cooperation is needed. In Composite Answer Situation
Case 3 and 4, we need transmit the complement concept to the service partners. The
Web CGI program makes a new Import on complement concept to partner server, as
we discussed in section 2. Moreover, we deliberately ignored the search of the actual
individuals (ABox) that satisfy a query, i.e. in the current implementation, only
TBoxes are considered.
Importer Web CGI Reasoning Processor Syntax Translator TBox
http://mediator/import
daml2al()
import()
classify()
Composite Answer Situation Case
CASC=4,5: make_import2partner(complement)
al2daml(satisfaction)
HTML page
Fig. 5 The UML Sequence of Import Process
4. Conclusion Remarks
The federated mediators is a heterogeneous environment, it means different
hardware and software, different knowledge representation systems. Thanks to W3C's
work, it is possible to communicate between heterogeneous hardware and software by
the family of standard from W3C (as XML, SOAP, WSDL, OWL, Web Services,
Semantic Web, etc.)
The Syntax Translator may translate the AL- concept object to multiple
knowledge representation languages in different message format. Which language
will be accepted by federated mediator server, DAML+OIL or OWL? Which
communication standards will be accepted, SMTP or HTTP? etc. The translator may
read the necessary information from mediator server's a description in WSDL.
Complete semantic interchange is a hard problem between heterogeneous
knowledge representation systems. In this work, we did not address semantic
translation between any knowledge representation languages. The Syntax translator is
a limited translator on semantic query, and only work at the syntax level. We see in
last section, the Complement Concept is under a conjunctive form: C = ( and C1,
C2, …, Cn). Each Ci is an atomic concept where each atomic concept is identified by a
term. The term is supposed to exist in a common lexical ontological dictionary (such
as WordNet[13], EuroWordNet[4]). Moreover, we suppose facilities for mapping
between terms. (Some term mapping approaches in lexical level are described [6]). In
our work, heterogeneous systems cooperation thanks to the complement concept
which is under the simplest form, conjunction. The conjunctive form is adopted in
most knowledge representation languages, so the complement concept may be
understood by the other partner systems. So these systems may be in different
knowledge representation languages.
From the experimental platform, we believe that heterogeneous systems may
interoperate by virtue of composite answer approach. We believe that the hard
problems do not go away even if we solve low-level issues such as natural language
analysis, complex term mapping, and ontologies integration. Further, we suggest that
the root causes can be understood better in terms of complement.
In the future, based on our initial experience on limited heterogeneous knowledge
representation technology and common mathematical background of heterogeneous
knowledge representation technology [21], two interrelated approaches paid attentions
to our work. The first is to find some limited translation approaches between any two
knowledge representation languages. The second is to measure the mismatch, then
describe it in a common knowledge structure, and find the algorithm to get the “best”
composite answer.
References
1. W3C: Semantic Web (2003) http://www.w3.org/2001/sw.
2. Rahm, E., Bernstein, P.A.: A survey of approaches to automatic schema matching. VLDB
Journal: Very Large Data Bases 10 (2001) 334–350
3. Garlan, D., Allen, R., Ockerbloom, J.: Architectural mismatch, or, why it’s hard to build
systems out of existing parts. In: Proceedings of the 17th International Conference on
Software Engineering, Seattle, Washington (1995) 179–185
4. Piek, V.: Eurowordnet: Amultilingual database with lexical semantic networks.
EuroWordNet General Document 3 (1999)
5. Ahrens, K., Chung, S.F., Huang, C.R.: Conceptual metaphors: Ontology-based rep-
resentation and corpora driven mapping principles. In John Barnden, Sheila Glas-bey,
M.L., Wallington, A., eds.: Proceedings of the ACL 2003 Workshop on the Lexicon and
Figurative Language. (2003) 35–41
6. Niles, I., Pease, A.: Linking lexicons and ontologies: Mapping wordnet to the suggested
upper merged ontology. In: Proceedings of the IEEE International Con-ference on
Information and Knowledge Engineering. (2003) 412–416
7. Boudjlida, N.: AMediator-Based Architecture for Capability Management. In Hamza, M.,
ed.: Proceedings of the 6th International Conference on Software En-gineering and
Applications, SEA 2002, MIT, Cambridge, MA (2002) 45–50
8. H.K. T¨onshoff H.K., I. Seilonen, G.T., P.Leitado: Amediator-based approach for
decentralised production planning, scheduling and monitoring. Proceedings of CIRP
International Seminar on Intelligent Computation in Manufacturing Engi-neering (2000)
113–118
9. Sudha Ram, J.P., Hwang, Y.: Cream: Amediator based environment for modelling and
accessing distributed information on the web. In: Nineteenth British Nantional
Conference on Database(BNCOD2002). (2002) 58–61
10. Preece, A.: Amediator-based infrastructure for virtual organisations. In Agents-2001
Workshop on Intelligent Agents in B2B E-Commerce (2001)
11. Cheng, D., Nacer, B.: Federated mediators for query composite answers. In: 6th
International Conference on Enterprise Information Systems -ICEIS’2004, Porto,
Portugal. Volume 4. (2004) 170–175
12. Pinto, H., Prez, A., Martins, J.: Some issues on ontology integration. In: Pro-ceedings of
the IJCAI-99 workshop on Ontologies and Problem-Solving Meth-ods(KRR5). (1999) In
[18], 7–1 – 7–12
13. Miller, G.A., al.: Wordnet 2.0 (2003) http://www.cogsci.princeton.edu/ wn/.
14. UDDI.org: Uddi: Universal description, discovery and integration. Technical White Paper
(2003) (http://uddi.org).
15. Beatty, J., al.: Web Services Dynamic Discovery (WS-Discovery) Specification.
Microsoft, BEA, Canon, Inc and Intel. (2004)
http://msdn.microsoft.com/ws/2004/02/discovery.
16. Horrocks, I.: DAML+OIL: a description logic for the semantic web. IEEE Data
Engineering Bulletin 25 (2002) 4–9
17. Randall Davis, H.S., Szolovits, P.: What is a knowledge representation? AI Mag-azine 14
(1993) 1993
18. DL-org: Description logics (2003) http//dl.kr.org/.
19. Kifer, M., Lausen, G., Wu, J.: Logical foundations of object-oriented and frame-based
languages. Journal of the Association for Computing Machinery 42 (1995) 741–843
20. Chein, M., Genest, D.: Cgs applications: Where are we 7 years after the first iccs? In
Ganter, B., W.Mineau, G., eds.: Conceptual Structures: Logical, Linguistic, and
Computational Issues. Volume 1867., Darmstadt, Germany, ICCS 2000, Springer (2000)
127–139
21. Sowa, J.F.: Mathematical background of knowledge representation (2003)
http://www.jfsowa.com/logic/math.htm.
22. Amati, G., Ounis, I.: Conceptual graphs and first order logic. The Computer Journal 43
(2000) 1–1