MEDI-ADAPT

A Distributed Architecture for Personalized Access

to Heterogeneous Semi-structured Data

Rim Zghal Rebaï, Amel Corinne Zayani, Ikram Ammous and Abdel Majid Ben Hamadou

Institute Multimedia, Information Systems and Advanced Computing Laboratory,

Higher Institute of Computer and Multimedia, PO Box 242, Sakiet Ezzit, 3021 Sfax, Tunisia

Keywords: AHA!, Navigation Adaptation, Content Adaptation, DARPA I3, Mediation, Semi-structured Data, XML,

RDF.

Abstract: The increased volume of data in digital form has led to a wide variety of syntactic and semantic data. Thus,

the user may encounter several problems related to the heterogeneity, distribution and volume of returned

information. As a matter of fact, systems are needed to solve part or all of these problems. In this paper we

propose and illustrate a distributed architecture that enables personalized access to collections of

heterogeneous and distributed semi-structured documents (XML, RDF, SMIL). This architecture is based on

an extension on the reference architecture of mediation systems DARPA I3 by adding an adaptation layer,

based on the reference architecture for adaptive hypermedia systems AHA!, that takes into account the

user’s request, his context (profile, device, network, etc.. ) and data sources context (minimum bandwidth,

necessary characteristics of the device to display data, etc.).

1 INTRODUCTION

Actually, data sources have become increasingly

heterogeneous and distributed all over the world. As

a result, the data volume grows and the user can not

access to relevant information not only to his needs

but also to his context. Thus, mediators are provided

to resolve the problem of access to these sources

regardless of their natures, semantics and locations.

But these mediators, like IRO-DB (Gardarin, 1995),

XMedia (Dang-Ngoc, 2008), (Kerzazi, 2008), etc.,

are unable to adapt the data and provide always the

same information to users despite differences in their

contexts (profile, device, network, etc.). This clearly

shows the need for systems able to deal with both

mediation and adaptation to provide a unified and

transparent access to distributed heterogeneous data

sources taking into account the user s’ needs and

context and the data sources’ context.

With the lack of this type of systems, we propose

an architecture of a distributed system mainly

establishing components for solving problems

related to remote and transparent access to document

collections, adaptation of data and context

management. The main objective of this architecture

is to offer the user an adaptation of the content and

navigation while accessing to collections of semi-

structured documents that are distributed,

syntactically and semantically heterogeneous and

not designed to be adapted.

The paper is organized as follows: In section 2,

we present a state of the art of some works dealing

with mediation and adaptation. In Section 3, we

introduce our proposal for a distributed architecture

treating jointly mediation and adaptation. We focus

in section 4 on the main layers of this architecture

and their components. We conclude and give

directions for future work in Section 5.

2 RELATED WORKS

The mediation systems allow the integration of

heterogeneous and distributed data sources. All the

proposed systems in the literature share a common

reference architecture DARPA I3 (Wiederhold,

1992). This architecture has three layers: client,

mediation and sources. From this reference

architecture, different generations of systems have

been developed based on the choice of pivot model

and language. The first mediation systems were

based on the relational model such as Multibase

259

Zghal Rebaï R., Corinne Zayani A., Ammous I. and Majid Ben Hamadou A..

MEDI-ADAPT - A Distributed Architecture for Personalized Access to Heterogeneous Semi-structured Data.

DOI: 10.5220/0003932402590263

In Proceedings of the 8th International Conference on Web Information Systems and Technologies (WEBIST-2012), pages 259-263

ISBN: 978-989-8565-08-2

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

(Landers, 1982), Mermaid (Templeton, 1987), etc.

In the 90s, a second generation of object-oriented

systems appeared such as IRO-DB (Gardarin, 1995),

GARLIC (Carey, 1995), DISCO (Tomasic, 1998).

Then, with the achieved success by XML (W3C,

1998), which is accepted as a standard for semi-

structured data, a generation based on this language

has emerged, such as XMedia (Dang-Ngoc, 2008).

However, with the advent of the Semantic Web

(Tim, 1999), XML has shown some limitations since

it can provide only descriptive metadata, unlike RDF

(Manola, 2004), which can provide both descriptive

and semantic metadata. That’s why, another RDF-

based generation of mediation systems emerged. The

advantage of such systems is that they take into

account the semantics of data as the proposed

systems in (Vdovjak, 2001), (Kerzazi, 2008).

Generally, when two users submit the same

request (by query or link), mediators provide the

same answers despite the difference of contexts.

This means that the mediators do not deal with

adaptation and do not take into account the context

of the user when accessing to sources.

In the literature, several models and architectures

dealing with adaptation are proposed. The first

reference model is “Dexter” (Halasz, 1994). It is a

local-oriented model proposed for hypertext

applications. An extension of this model is proposed

in (De Bra, 1999) called AHAM (Adaptive

Hypermedia Application Model). This latter was

extended in the local context to take into account

semi-structured XML documents (Zayani, 2008).

The emergence of client / server architecture has

led to the development of AHA! (Adaptive

Hypermedia Architecture) (De Bra, 2003). It is a

web-oriented architecture based on AHAM and

considered as a reference architecture for adaptive

hypermedia systems. Among the architectures based

on AHA! we cite CA-WIS (Soukkarieh, 2010)

which is proposed to take into account web services.

Another type of mediator-oriented architecture is

proposed to allow the resolution of adaptation and

mediation problems together. The only one is

proposed in (Kostadinov, 2008). It offers a

personalized access to many distributed relational

data sources and uses XML as a pivot language in

the mediator level. It adapts the content taking into

account the user’s profile and the available sources’

quality. But in this work: (i) all of data are relational;

(ii) the syntactic heterogeneity of these data is

supposedly resolved; (iii) the semantic heterogeneity

is neglected; and (iv) the navigation adaptation is not

treated.

We notice that all the already mentioned studies

adapt homogenous and known in advance data,

except (Zayani, 2008) which adapts unknown local

semi-structured data. Moreover, the navigation

adaptation is not treated by the majority of these

studies despite its importance, especially in the case

of a large amount of data like in distributed

environments. In these environments, the user can be

easily disoriented and cannot get the required

information, thus the benefits of navigation and

content adaptation are more visible and efficient.

That’s why we propose a mediator-oriented

architecture able to deal with; (i) the navigation and

content adaptation, (ii) the access to semantic and

syntactic heterogeneous and distributed semi-

structured data (XML, RDF, SMIL) by using RDF

as a pivot language in the mediator; and (iii) takes

into account the variety in the contexts of users and

sources.

3 THE MEDI-ADAPT

ARCHITECTURE

We propose a distributed architecture that ensures

the adaptation of content and navigation while

accessing to heterogeneous sources not designed to

be adapted.

It is considered to be innovative because, it

executes multiple functions; it (i) deals with the

navigation and content adaptation in a distributed

environment; (ii) uses a mediator that provides

transparent access to collections of semi-structured

documents; (iii) treats syntactic and semantic

heterogeneity of sources; (iv) takes into account both

the user’s context and the data sources’ context.

This architecture is essentially based on a

combination of DARPA I3 (reference architecture

for mediation systems) and AHA! (reference

architecture for hypermedia adaptive systems). We

extended the DARPA architecture by adding an

adaptation-providing layer. This layer is essentially

based on the AHA!.

The architecture that we propose is consequently

composed of four layers: client, adaptation,

mediation and sources. As it is illustrated in Figure

The client layer allows the user-system

interaction. Thus, the user makes his access (by

query or link) using various devices (PC, PDA, cell

phone, etc.).

The adaptation layer receives the user’s demand

from the client layer and adapts the content and

navigation according to the user's needs and context

WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies

260

as well as the sources’ context.

The mediation layer provides a unified and

transparent access to different sources, regardless

their syntax, semantics and location.

The sources layer integrates several collections

of semi-structured heterogeneous and distributed

documents.

Figure 1: The MEDI-ADAPT architecture.

4 THE DESCRIPTION OF THE

MAIN PROPOSED LAYERS

The main layers of the proposed architecture are the

adaptation layer and the mediation layer. These two

layers collaborate together to provide the content

adaptation (Peter Brusilovsky,1996), the

personalized access to the sources layer and the

navigation adaptation (Peter Brusilovsky,1996)

taking into account different contexts.

We detail below their components and their

performance.

4.1 The Adaptation Layer

In the adaptation layer there are: a User Context

Manager (UCM), a Query Interpreter (QI), a Content

Adaptation Engine (CAE), an Updating Profile

engine (UPE), a Navigation Adaptation Engine

(NAE), a Response Generator (RG) and a

Navigation Manager (NM).

This layer is illustrated in Figure 2.

Each component of this layer provides a well

defined role:

 The UCM manages all components of the user's

context: profile and environment (network and

device). It is composed of a Profile Manager

(PM), a User Model (UM) and an

Environment Manager (EM). The UM stores

the user’s profile as proposed in (AHA!,

(Zayani,2008), CA-WIS). We suggest that the

user’s profile is divided into two parts: one

contains data concerned with content

adaptation and another one contains data

concerned with navigation adaptation;

 The CAE and the UPE, respectively, adapts the

content and updates the user’s profile. They

are similar to the downstream adaptation

mechanism and the updating profile

mechanism proposed in (Zayani, 2008);

 The NAE adapts the navigation. It is based on

an extension of the upstream adaptation

engine proposed in (Zayani, 2008) by tacking

into account the navigation adaptation;

 The RG performs the transformations generated

by the NAE on the results data before being

displayed to the user as it is the case with

FAWIS (De Virgilio, 2007) and CA-WIS;

 The QI receives the user’s request, analyses and

transforms it into a SPARQL

(Prud'hommeaux, 2008

) (the W3C

Recommendation for an RDF query language)

query that is understandable to all the system

components.

Figure 2: The adaptation layer’s components.

When the user launches a request to the system,

the UCM treats four actions simultaneously: by the

PM, it (i) recognizes the user (ii) looks for his profile

from the UM (iii) analyses it in order to differentiate

between its two parts (content and navigation) and

by the EM, it detects the user's environment (the

MEDI-ADAPT-ADistributedArchitectureforPersonalizedAccesstoHeterogeneousSemi-structuredData

261

used device and the network’s characteristics). The

QI receives the user’s request, analyses and

transforms it into a SPARQL query. Then, the CAE

expands this query with the extracted data from the

content part of the profile and sends it to the

mediator. The latter brings the documents results

from sources and sends them to the NAE. This

engine receives from the UCM the navigation part of

the profile and the user’s environment characteristics

to make the navigation adaptation and sends the

result to the RG which, in its turn, generates the final

result and sends it to the user.

Throughout each session, the NM detects the

user-system interactions (visited links, accessed

documents, duration of each document consultation,

session’s length, etc.) and sends them to the PUE.

4.2 The Mediation Layer

The mediation layer is mainly based on the basic

components of XMedia (Tuyet et al., 2008): a

Metadata Manager (MM), a Query Decomposer

(QD), a Query Executor (QE), a Result

Reconstructor (RC), a Metadata Base (MB) and a

Memory Cache (MC). Except the Query Interpreter

(QI) which is moved, as it is mentioned above, to the

adaptation layer.

This layer is illustrated in Figure 3.

Figure 3: The mediation layer’s components.

The components of this layer work as follows:

 The Metadata Manager (MM) provides

descriptive metadata for each documents’

collection, interprets and converts them to the

pivot language RDF.

 The Query Decomposer (QD) decomposes the

SPARQL query into atomic queries according

to each single collection. It identifies the

appropriate metadata sources, locates the

collections and creates a query execution plan.

 The Query Executor (QE) executes the atomic

queries near the sources, receives the results

and sends them to the MC and the RC.

 The Result Reconstructor (RC) gathers all the

returned results and sends them to the NAE.

 The Metadata Base (MB) stores the metadata

related to documents’ collections.

 The Memory Cache (MC) stores all the

submitted data to the adaptation layer during a

session.

When the CAE (component of the adaptation

layer) sends a query to the mediation layer, the QD

receives and decomposes it into atomic queries

according to each single collection. It identifies the

appropriate metadata sources, locates the collections

and creates a query execution plan. The QE receives

all the atomic queries and executes them near the

sources, then it receives the results from there and

sends them to the MC and the RC. The latter gathers

all the returned results and sends it to the NAE

which performs the navigation adaptation task as it

is mentioned above. Finally, the RG sends the final

result to the user.

Permanently, the MM connects with different

sources to provide descriptive metadata for each

documents’ collection, converts all of them to RDF

to be stored in MB and identifies the global schema.

5 CONCLUSIONS

In this paper, we presented a distributed architecture

that solves both mediation and adaptation problems

together. It is essentially based on a combination of

DARPA I3 (reference architecture for mediation

systems) and AHA! (Reference architecture for

hypermedia adaptive systems).

On the one hand, this architecture offers a

unified and transparent access to distributed

collections of semi-structured documents through

the mediation layer. On the other hand, it offers a

navigation and content adaptation taking into

account the user and sources contexts through the

adaptation layer. The syntactic and semantic

heterogeneity of data is solved by the use of RDF as

a pivot model at the mediation layer.

There are many perspectives that we are willing

to realize. First, we plan to propose and evaluate a

navigation adaptation method on which the

Navigation Adaption Engine (NAE) will be based

(already in the evaluation step). Second, we aim to

WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies

262

propose a content adaptation method on which the

Content Adaptation Engine (CAE) will be based.

Third, we will suggest methods that treat the data

heterogeneities (syntactic and semantic) and

distribution. Also, we intend to propose, at the

Profile Manager (PM), a method that reduces the

user profile to the most relevant content. Finally, we

are going to implement a prototype based on the

proposed architecture.

REFERENCES

Brusilovsky, P., 1996. Methods and techniques of adaptive

hypermedia. User Modeling and User Adapted

Interaction, 6(2-3):87–129.

Carey, M., Haas, L., Scwartz, P., 1995. Towards

Heterogeneous Multimedia Information Systems: The

Garlic Approach, in RIDE-DOM, pages 124-131, 19.

Dang-Ngoc, T., Gardarin, G., 2008. Conception et

Evaluation de XQuery dans une architecture de

médiation “Tout-XML”. CoRR abs/0806.4920.

De Bra, P., Geert-Jan, H., Hongjing, W., 1999. AHAM: A

Dexter-based Reference Model for Adaptive

Hypermedia. Hypertext 99 Darmstadt Germany.

De Bra, P., Aerts, A., Berden, B., De Lange, B., Rousseau,

B., Santic, T., Smits, D., Stash, N., 2003. AHA! The

Adaptive Hypermedia Architecture. HT’03.

De Virgilio, R., Torlone, R., Houben, G., 2007. Rule-

based Adaptation of Web Information Systems. World

Wide Web.

Gardarin, G., Fankhauser, P., Finance, B., Klas, W.,

Ramfos, A., Gannouni, S., Pastre, D., Legoff, R.,

1995. IRO-DB: A Distributed System Federating

Object and Relational Databases. Object-oriented

Multibase Systems, Prentice-Hall.

Halasz, F., Schwartz, M., 1994. The dexter hypertext

reference model. ACM, 37(2):30–39.

Kerzazi, A., Chniber, O., Navas-Delgado, I., F.Aldana-

Montes, J., 2008. A Semantic Mediation Architecture

for RDF Data Integration. SWAP’08.

Kostadinov, D., Bouzeghoub, M., Lopes, S., 2008. Accès

personnalisé à des sources de données multiples.

Ingénierie des Systèmes d'Information 13(4): 59-82.

Landers, T., Rosenberg, R. L., 1982. An overview of

MULTIBASE. In distributed databases, H.J Shneider

editor, North-holland.

Manola. Frank., Milles, E., 2004. RDF Primer. W3C

Recommendation, http://www.w3.org/RDF/.

Prud'hommeaux, E., Seaborne, A. (eds), 2008. SPARQL

Query Language for RDF. W3C Recommendation,

http://www.w3.org/TR/rdf-sparql-query/.

Soukkariah, B., 2010. Technique de l’Internet et ses

langages: Vers un Système d’Information Web

restituant des Services Web sensibles au contexte

Doctoral thesis, University of Toulouse 3 – Paul

Sabatier, Toulouse, France.

Templeton, M., Brill, D., Dao, S. K., Lund, E., Ward, P.,

Chen, A. L., Macgregor, R., 1987. Mermaid: a

Frontend to Distributed Heterogeneous Databases. In

proceedings of the IEEE conference on Data

Engineering.

Tim, B., Mark, F., 1999. Weaving the web, chapter

Machines and the Web, pages 177 – 198. Harper, San

Francisco.

Tomasic, A., Raschid, L., Valduriez, P., 1992. Scaling

Access to Heterogeneous Data Sources with DISCO,

IEEE Data and knowledge Engeneering, 10(5):8, No.

3, 3849.

Vdovjak, R., Houben, G., 2001.RDF Based Architecture

for Semantic Integration of Heterogeneous

Information Sources, In Workshop on Information

Integration on the Web, pp. 51-57.

Wiederhold, G., 1992. Mediators in the architecture of

future information systems. IEEE Computer

Magazine, Vol. 25, No. 3, 3849.

W3C, 1998. EXtensible Markup Language (XML) 1.0.

Technical report, World Wide Web Consortium

(W3C), Technical report, February.

Zayani, A., 2008. Contribution à la définition et à la mise

en œuvre de mécanismes d’adaptation de documents

semi-structurés. Doctoral thesis, University of

Toulouse 3 – Paul Sabatier, Toulouse, France.

MEDI-ADAPT-ADistributedArchitectureforPersonalizedAccesstoHeterogeneousSemi-structuredData

263