ADAPTIVE INTEGRATION OF INFORMATION

Christophe Nicolle and Christophe Cruz

Laboratory Le2i, Université de Bourgogne, B.P. 47870, 21078, Dijon CEDEX, France

Keywords: XML, Data integration, OWL, RDF, CDMF.

Abstract: This paper suggests a new approach to improve the integration of information. This approach is based on

the use of semantic adaptive graphs. The adaptive feature of our proposal makes it possible to manage two

specific aspects related to information integration: the adaptation of information according to the user’s

access rights and the lifecycle of the integrated information.

1 INTRODUCTION

In the field of information systems, integration

consists in unifying heterogeneous information

sources for a given user. Since the Seventies, the

scientific community has tried to build up

this unified view through the proposal of various

models, languages and architectures independently

of an implementation (Navathe, 1982), (Chen,

1976). Today, the integration of information

becomes a set of calls made to Web services, which

are integrated into an XML document (Abiteboul,

2008). The resulting graph is built dynamically

according to the results of the distant calls.

Nevertheless, for each proposal, semantic

heterogeneity remains an unsolved subjacent

problem. In this field, the interest is to model a tacit

knowledge which is related to the information. The

suggestions evolve from the models to the

metamodels, then towards the mediators and finally

to ontologies (Liu, 2007). During this period, the

structure is strongly influenced by the emergence of

the object oriented models. Then, derived from the

concept of graphs (Sowa, 1984), the concept of the

semantic graph appears, as well as the concept of

ontology (Guarino, 1994). In our example, the

integration of information using ontology makes it

possible to assemble two pieces of a puzzle

according to the image which is printed on them

without worrying about the shape of the pieces.

From the combination of XML, ontology and

Web services, new proposals appear. For the

structural part, the comparison of XML grammars

becomes an important field of study. These

proposals are the descendants of those carried out by

Miller developed in 1993 (Miller, 1993). For the part

that is concerned with information access,

orchestration and choreography languages are

developed to combine the Web services. For the

semantical part, OWL and RDF languages become

the angular stones of research on ontologies. The

combinations of ontologies and XML are effective

when they are used to integrate information with the

objective of building a global system within the

meaning of the first distributed architectures (Bell,

1992), i.e. the source systems will disappear. Only

the built target system will be used. The formats of

the source models are replaced by XML schema.

The heterogeneous semantics of distributed

information will be homogenized in a common

thesaurus defined by ontology.

In the industrial field, the lifecycle of information is

of fundamental importance in the integration

process. This process is conditioned by the nature of

the information but also by the way in which it is

used. The nature of information can change

according to the context of use. In this approach,

integration is not finality but it is a continuous

process. During this process it is possible to

dynamically adapt the integrated information

according to the lifecycle of local information from

data source systems. To reach these requirements,

this paper presents a new approach based on

adaptive semantic graphs. These graphs model

information as well as the contexts of use of this

information and the lifecycle of this information for

each context. Our model is based on a combination

of various operators, such as RDF, OWL, SWRL

and Named Graph. The nodes described in these

115

Nicolle C. and Cruz C.

ADAPTIVE INTEGRATION OF INFORMATION.

DOI: 10.5220/0001846701150118

In Proceedings of the Fifth International Conference on Web Information Systems and Technologies (WEBIST 2009), page

ISBN: 978-989-8111-81-4

graphs can represent multi-media information,

operators on the graphs and sub-graphs.

The following section is made up of three parts.

The first part gives an overview of the framework

CDMF. The second part presents a brief description

of the operators used to model the structural part of

the graph. The third part presents the operators used

to model the contextual part and the lifecycle of

information. The last part concludes this paper.

2 CDMF OVERVIEW

The CDMF framework gives a set of operators and

modeling structures which allow to deal with

temporal and contextual requirements of the

adaptive integration of information. The applicative

environment developed for CDMF is composed of

several parts: the engine, the configuration graph

called “SpaceSystem” and the interface (called API

CDMF).

 The engine is composed of the Data Modeling

Layer and the Context Modeling Layer. The first

layer relates to the inference engine which is

used to infer and to check the data modeled by

the CDMF modeling operators. It is made up of

the implementation of the graph combination

operators (such as AddGraph, RemoveGraph,

MapGraph, etc.) and of the element

SystemGraph. The second layer is used to

manage the contexts in the CDMF architecture.

 The SpaceSystem constitutes a configuration

space used by the inference engine. This space

contains a set of graphs SystemGraph containing

the initialization data.

 The API CDMF proposes a set of functions

giving access to the various functionalities of the

CDMF applicative environment.

Due to lack of space, we restrict our presentation to

the Data Modeling Layer (DML) and the Context

Modeling Layer (CML). DML defines a reduced set

of operators allowing the semantic modeling of

information. These operators are based on RDF

specifications. CML is dedicated to context

modeling and graph manipulation.

3 THE DATA MODELING LAYER

DML is a language composed of a set of operators

derived from RDF, OWL and SWRL. DMF makes it

possible to describe classes and properties. These

classes and these properties can then be used in

statements (formulas) using operators of implication,

intersection, union, etc. The operator of implication

allows to constitute rules which express constraints

on these sets of individuals. The operators

composing this layer are



dmf:Class defines a class.



dmf:Property defines a property.



dmf:Equal defines the equality of two resources.



dmf:Var defines variables used in the logical

formulas.



dmf:Predu defines unary predicates.



dmf:Predb defines binary predicates.



dmf:Equiv defines two predicates as equivalent.



dmf:And defines the intersection.



dmf:Not defines the negation.



dmf:Or defines the union.



dmf:OrX defines the exclusive disjunction.

For the following statements, the operator of

implication

dmf:Imp is used to represent various

operators used in OWL.

dmf:Imp is equivalent to the

operator

ruleml:Imp defined in SWRL. This SWRL

operator is derived from the RuleML formalism.



Imp(p1(?x,?y),p2(?x,?y)): defines p1 as a

sub-property of p2.



Imp(p(?x,?y),And(A(?x),B(?y)): defines

restrictions of the type for the subject and the

object of a property. Here, all the subject

instances of the property p are of type A and all

the object instances of the property p are of type



Imp(And(p(?x,?y),p(?y,?z)),p(?x,?z)):

defines a transitive property



Imp(p(?x,?y),p(?y,?x)): defines a

symmetrical property.



Imp(And(p(?x,?y),p(?x,?z)),Equal(?y,?z)):

defines a functional property. This feature

indicates a single property. It is a short cut to

declare a minimal cardinality being 0 and a

maximal cardinality being 1.



Imp(p1(?x,?y),p2(?y,?x)): defines p2 like the

inverse property of p1.



Imp(A(?x),B(?x)): defines A as a subclass of B.



Imp(And(p(?x,?y),p(?z,?y)),Equal(?x,?z)):

defines an inverse functional property.



Imp(And(A(?x),p(?x,?y)),B(?y)): defines all

the values of p as being of type B.

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

116



Imp(A(?x),[i,j]p(?x,?y)): defines

cardinalities which are defined by a couple of

values between square brackets. The first value

defines the minimal cardinality and the second

value defines the maximum cardinality. In this

statement, any element of the type A contains

properties from i to j.



Imp(A(?x),p(?x,value)): defines a default

value for a class and its property. The elements

of the type A have a property p whose value is

“value”.

The following example illustrates in an RDF/XML

format the use of the DML operators for the

construction of a schema. The two first lines create a

class

Building and a class Room. The third line

creates a property

contains. This example

composes the

cdmf:model which is part of the

dmf:systemgraph presented in the next section.

1 <dmf:Class rdf:ID=’Building’/>

2 <dmf:Class rdf:ID=’Room’/>

3 <dmf:Property rdf:ID=’contains’/>

The next example presents a DMF graph modeling

data according to the previous schema. Due to lack

of space, the format is in RDF/N3. This example

composes the

cdmf:graph which is part of the

dmf:systemgraph presented in the next section.

1 :G1 {

2 :room_1 rdf:type :Room

3 :building_1 rdf:type :Building

4 :building_1 :contains :room_1 }

The first line defines a data graph called G1. Line 2

defines two instances of class

Room called room_1.

Line 3 defines an instance

building_1 from the

class

Building. This instance is linked with room_1

using the property

contains in line 4. This example

presents the static part of our approach. This part

makes it possible to integrate heterogeneous

information without taking into account the lifecycle

of information.

4 CML

The Context Modeliing Layer (CML) has been

developed to enhance the static part defined in the

DML. CML is articulated in two parts. The first part

defines the context of use for each DML graph. The

second part defines a set of operators used to

combine graphs and to limit data redundancy.

CDMF organizes a set of properties into a graph.

These properties are used to model control access

(read/write) and to represent context. This element is

called

cdmf:SystemGraph. It is composed of

properties that describe with the help of graphs the

user’s access conditions, the data model in which the

data graph is structured and the data graph:



cdmf:graph connects graph and data. These data

are described according to the data model (which

can be a combination between other graphs using

CDMF operators).



cdmf:of represents the context. This property

defines a list of resources representing the access

context.



cdmf:model defines for a system graph the data

model which is used. This data model defines

elements which will appear in the graph.



cdmf:action defines user’s rights to access the

data. (read/write/remove)

We derive from Named Graph the representation of

contexts. This property makes it possible to define

an RDF graph like a resource related to a context.

The sub-graph modeling the context is defined by

the property

cdmf:of. In our proposal we used the

context to construct an integrated sub-system

according to the type of user. It enables us to build

an integrated information management according to

user environments. For example, it is possible to link

the data graph G1 with a context defined by a user

(Line 1) and a date (Line 2).

1 :G1 :auteur ’Christophe Cruz’.

2 :G1 :date ’08/23/08’.

The second part of CML is made of operators. The

use of these operators allows the simplification of

the management of the evolution of integrated

information. Rather than to store a new version of

information, we integrate the process of evolution of

the information into a graph of operators.



cdmf:AddGraph allows the union of two or

several graphs. It has a property

cdmf:args. This

property is a list of RDF elements (

rdf:Bag)

which are graphs.

To illustrate this operator, let us consider the first

example G1 with a Building_1 which contains

room_1. If the next day ‘08/24/08’, the same user

‘Christophe Cruz’ updates this structure by adding a

new room to the building, the following new

cdmf:SystemGraph G2 will be built. The

cdmf:graph part will contain:

1 :G2 rdf:type cdmf:AddGraph

2 :G2 cdmf:args :li1

3 :li1 rdf:li :G1

4 :li1 rdf:li :G1b

5 :G1b {

6 :room_2 rdf:type :Room.

7 :building_1 :contains :room_2 }

ADAPTIVE INTEGRATION OF INFORMATION

117

The cdmf:of part contains the reference to the

context:

1 :G2 :auteur ’Christophe Cruz’.

2 :G2 :date ’08/24/08’.

The updates of the integrated information are

stored in the graph. In this example, it is possible to

go back to a former version of the integrated system.



cdmf:InterGraph carries out the intersection on

the sets of triplets of each graph. It has two

properties

cdmf:arg1 and cdmf:arg2, which

represent the two graphs on which the

intersection must be calculated



cdmf:CompInterGraph makes it possible to

determine which part of the triplet is concerned

by the calculation of the intersection.



cdmf:MapGraph defines a graph of

correspondence. This graph is a transformation

of a graph into another graph using rules of

correspondences. It has two properties

cdmf:src

and

cdmf:map indicating the graph source and the

graph where the rules of transformation are

defined.



cdmf:RemoveGraph makes it possible to remove a

part of a graph. It has two properties

cdmf:src

and

cdmf:rem. The second property constitutes

the set of the triplets to be withdrawn from the

graph indicated by the first argument.

We tested several combinations of operators in order

to:

 preserve the history of the integrated information

updates

 adapt the model of integrated data according to

the new source models which are added during

the lifecycle of the system.

 ensure the migration of the data according to the

evolution of the integrated data model.

 build interfaces adapted to the profile of the user.

An interface is a sub-graph which connects the data,

the process and a graphic charter according to the

rights and the context of the user.

5 CONCLUSIONS

This paper made an attempt at presenting a state-of-

the-art related to the integration of information in the

field of information systems. The authors tried to

underline the difference between the academic

proposals and the actual industrial realities. The

paper showed that integration is not a final process.

The integration of information in the industrial

world requires a perpetual update of the integrated

system while taking into account the lifecycle of

information and the user context. To meet these

needs, this paper presented an integration method

based on adaptive semantic graphs. These graphs

make it possible to facilitate the process of

integration while proposing mechanisms to update

the integrated information and its context

representation.

Our proposal was implemented into a Web

collaborative platform dedicated to facility

management. In this field, the lifecycle of the

building is made up of four steps: design,

construction, exploitation and maintenance. In this

platform, the local users preserve the use of their

local systems. Each local user can obtain an

integrated sight of the building in the form of a 3D

mock-up generated from the operators of CDMF.

Currently, more than 6 million m

of building

surface are being integrated.

REFERENCES

Abiteboul, S., Benjelloun, O., Milo, T., The Active XML

project: an overview, VLDB Journal, 2008.

Bell, D., Grimson, J., Distributed Database Systems, Int.

Computer Science Series, Addison-Wesley Publishing

Company, 1992.

Chen P.P., The Entity-Relationship Model Toward a

Unified View of Data, ACM Transaction on Database

Systems, Vol.1, N°1, 1976.

Guarino, N., Carrara, M., Giaretta, P., An Ontology of

Meta-Level Categories. Proc. of the 4th Int. Conf. on

Principles of Knowledge Representation and

Reasoning, pp 270-280 Bonn, Germany, 1994.

Miller, R.J., Ionnadis, T.E, Ramkrishnan, R., The Use of

Information Capacity in Heterogeneous Systems:

Bridging Theory and Practice, Proc. of the 19th

VLDB Conf., Dublin, Ireland, 1993

Liu, Q., Huang, T., Liu S., Zhong, H., An Ontology-Based

Approach for Semantic Conflict Resolution in

Database Integration. Journal of Computer Science.

Technology Vol. 22 N°2, pp 218-227, 2007.

Navathe, S.B., Gadgil, S.G., A, Methodology for View

Integration in Logical Database Design, Proc. of the

8th Inter. Conf. on VLDB, pp142-164, 1982.

Sowa, J.F., Conceptual Structures: Information

Processing in Mind and Machine, Addison-Wesley

1984.

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

118