ONTOLOGY-BASED MEDIATION OF OGC CATALOGUE SERVICE

FOR THE WEB

A Virtual Solution for Integrating Coastal Web Atlases

Yassine Lassoued

, Dawn Wright

, Luis Bermudez

and Omar Boucelma

Coastal and Marine Resources Centre, University College Cork, Naval Base - Haulbowline, Cobh - Co. Cork, Ireland

Department of Geosciences, Oregon State University, Corvallis, Oregon, U.S.A.

Southeastern Universities Research Association (SURA), Washington DC, U.S.A.

Laboratoire des Sciences de l’Information et des Systmes

Avenue Escadrille Normandie Niemen, 13397 Marseille, France

Keywords:

Data Semantics, Semantic Web Technologies, Information retrieval, Geographic Information Systems (GIS),

Ontologies, Catalogue Service for the Web (CSW), Mediation.

Abstract:

In recent years signiﬁcant momentum has occurred in the development of Internet resources for decision

makers and scientists interested in the coast. Chief among these has been the development of coastal web

atlases (CWAs). While multiple beneﬁts are derived from these tailor-made atlases (e.g., speedy access to

multiple sources of coastal data and information), the potential exists to derive added value from the integration

of disparate CWAs, to optimize decision making at a variety of levels and across themes. This paper describes

the development of a semantic mediator prototype to provide a common access point to coastal data, maps and

information from distributed CWAs. The prototype showcases how ontologies and ontology mappings can be

used to integrate different heterogeneous and autonomous atlases, using the Open Geospatial Consortium’s

Catalogue Services for the Web.

1 INTRODUCTION

The vast and heterogeneous amount of geospatial data

on the World Wide Web causes users to be informa-

tion overloaded (Kashyap and Sheth, 2000). Search

engines return millions of results, including non rel-

evant information, that are rarely taken into account.

A user (e.g. scientist or a coastal response manager)

would like to have as much relevant information as

possible integrated for a particular event and region.

This requires 1) providing a mechanism to allow uni-

ﬁed discovery and access to distributed and heteroge-

neous data and 2) categorizing the results in a conve-

nient vocabulary to the end user.

Practically, data discovery relies on documenta-

tion provided as part of metadata (notably discovery

metadata) such as the dataset title, abstract, extent,

keywords, etc. In the context of distributed resources,

this information is present in different heterogeneous

formats, and systems, according to several existing

metadata models and standards.

The International Standardization Organization

(ISO) has recently deﬁned metadata standards for

geospatial data, notably the ISO-19139 (ISO, 2006)

standard. The aim is to harmonize metadata represen-

tation and implementation by conforming to a uniﬁed

model, a uniﬁed structure and a uniﬁed format. The

Open Geospatial Consortium (OGC), in their turn, fo-

cus on developing standards for querying and trans-

porting data and metadata over the Internet. Speciﬁ-

cally, the OGC Catalogue Service for the Web (CSW)

(Nebert and Whiteside, 2005) speciﬁcation deﬁnes a

standard for advanced querying and transporting of

metadata records over the Web.

Despite these harmonization efforts, problems still

arise when dealing with metadata semantics. Ter-

minology used to describe similar data can vary be-

tween specialities or regions, which can further com-

plicate data searches and integration. For instance,

usage of of the word ”seabed” in Europe versus use

of the word ”seaﬂoor” to describe the same feature

in North America is a good example of this scenario,

as is the interchangeable use of ”coastline” versus

”shoreline” in both regions. From both human and

computational standpoints, users need assurance that

the concepts, terminology, and even the abbreviations

that are shared between two or more individuals, sys-

tems, or organizations are understood by all to mean

192

Lassoued Y., Wright D., Ber mudez L. and Boucelma O. (2008).

ONTOLOGY-BASED MEDIATION OF OGC CATALOGUE SERVICE FOR THE WEB - A Virtual Solution for Integrating Coastal Web Atlases.

In Proceedings of the Third International Conference on Software and Data Technologies - ISDM/ABF, pages 192-197

DOI: 10.5220/0001891001920197

 SciTePress

the same thing. In this way the quality of data retrieval

and subsequent data integration are greatly increased.

In this paper, we describe an ontology-based

mediation approach for performing geospatial data

search across different organizations. We use ontolo-

gies as the means to deﬁne semantics for metadata

values (terms such as keywords, places, etc.) within

organizations, but also to link terms from different or-

ganizations. An organization or a group of organi-

zations populate their metadata using a local CSW.

Their metadata records use a given ontology of terms

called local ontology. Human or machine users for-

mulate CSW requests using a common ontology of

metadata terms, called global ontology. A CSW me-

diator rewrites the user’s request into CSW requests

over local CSWs using their own (local) ontologies,

collects the results and sends them back to the user.

The paper is organized as follows. In section 2

we sketch the problem through a concrete integration

example, while in section 3, we present state of the

art interoperability techniques and technologies. In

Section 4 we detail our CSW mediation approach, and

in section 5 we present the implemented prototype.

Finally, we conclude in section 6.

2 MOTIVATING EXAMPLE

The example described in this article is drawn from

the coastal web atlases (CWAs) integration problem

addressed by a new International Coastal Atlas Net-

work (ICAN) initiative (Wright et al., 2007).

A CWA is a Web geographic information system

composed primarily of coastal GIS data (vector data,

coverages, raster grids, and images), their associated

metadata, and thematic information about data (such

as textual descriptions, references, images, etc.).

Integrating several CWAs requires three different

levels of integration: integrating the GIS data, inte-

grating their metadata, and integrating the thematic

information that accompany data. The problem we

are focusing on in this article is metadata interoper-

ability, in other terms performing data discovery and

search across different CWAs in a transparent way.

We report here on the development of a prototype

as a proof-of-conceptto inter-relate metadata between

two initial CWAs: the Marine Irish Digital Atlas or

MIDA, [http://mida.ucc.ie], and the Oregon Coastal

Atlas or OCA, [http://www.coastalatlas.net]. It may

not be immediately obvious how Oregon and Ireland

may need to be interoperable, but these two mature

atlas efforts can be used as a testbed for interoperabil-

ity.

Both MIDA and OCA atlases implement the OGC

CSW for querying and delivering metadata records.

Metadata records in each of the CSWs use a given

local ontology (e.g. keywords, places, titles, etc.). In

order to facilitate search across both atlases, a central-

ized system needs to be implemented, which will pro-

vide uniﬁed and transparent access to the local CSWs.

Ideally, metadata records from both MIDA and OCA

will not be copied at the integrated level as both at-

lases are autonomous and are subject to evolution.

Rather the integration system will act as a mediator

(Wiederhold, 1992) that uses a common terminology

(metadata ontology) and will translate user queries,

on the ﬂy, into queries over the atlases’ CSWs using

their own ontologies.

3 STATE OF THE ART

This section presents the state of the art of existing

techniques and technologies that are related to the

problem of semantic interoperability of GIS metadata

management systems. The following subsections will

explore the use of CSW to facilitate syntactic interop-

erability (c.f. subsection 3.1) and the use of media-

tion as integration approach (c.f. subsection 3.2).

3.1 OGC Catalogue Service for the Web

(CSW)

CSW (Nebert and Whiteside, 2005) is an OGC ab-

stract speciﬁcation for supporting the ability to pub-

lish and search collections of descriptive information

(metadata) for data, services, and related information

objects. CSWs allow a uniﬁed access to metadata

records within a community or an organization, thus

harmonizing GIS resources discovery and search. For

this reason, CSWs are required for coastal atlases in

order to facilitate syntactic and schematic interoper-

ability.

CSWs support several operations. We focus here

on the

tt GetRecords operation for searching metadata

records, possibly using ﬁlters, such as keyword, lo-

cation and time search.

There exist several implementations of CSW. For

instance, both MIDA and OCA use GeoNetwork

[http://geonetwork-opensource.org] as a CSW imple-

mentation.

3.2 Mediation

The database (DB) community has extensively stud-

ied and developed data integration approaches and

systems leading to, among others, a virtual approach

ONTOLOGY-BASED MEDIATION OF OGC CATALOGUE SERVICE FOR THE WEB - A Virtual Solution for

Integrating Coastal Web Atlases

193

to data integration called mediation (Wiederhold,

1992).

A mediation system provides the user with a uni-

form interface of the different data sources via a com-

mon model. In a typical mediation architecture (c.f.

Figure 1), several distributed data sources use their

own data schemas, called local schemas or source

schemas. Users pose queries over a common refer-

ence schema, called global schema. The mediator

uses mapping rules between the global schema and

the local schemas in order to rewrite the user’s query

into queries over the local data sources. It extracts and

reformulates the responses conforming to the global

schema and combines them in order to construct a re-

sponse which is as complete as possible. Mediators

often use a wrapper for each data source, for translat-

ing queries from the mediator’s query language into

the data source’s query language, and for converting

the source’s data into the mediator’s data model. All

is transparent to the user. That is, the user ignores

where and how data are stored and how the mediator

manages to retrieve data from their sources.

Legend

Query flow

Data flow

Source nSource 2Source 1

Mapping Rules

Global Schema

Mediator

Wrapper Wrapper Wrapper

Figure 1: Typical Mediation Architecture.

Several mediation systems and prototypes have

been developed: examples of such systems are On-

tomet (Bermudez, 2004), TSIMMIS (Garcia-Molina

et al., 1997), PICSEL (Goasdoue et al., 2000), In-

formation Manifold (Kirk et al., 1995), AGORA

(Manolescu et al., 2001). Most of these existing

mediation approaches focus on data schemas hetero-

geneity and data complementarity and try to build re-

sponses which are as complete as possible.

4 CSW ONTOLOGY-BASED

MEDIATION

The targeted integrated Coastal Web Atlas, called

global

atlas or super atlas, is a virtual atlas that of-

Please note that the term ”global” does not refer to the globe in

this context. Rather, it is the term used by the database community

fers transparent access to a variety of distributed and

heterogeneous local coastal atlases. The notion of

”virtual”, in this context, means that local atlas re-

sources are not integrated or copied at the integrated

level. Rather, they remain at their locations and are

remotely accessed, harmonized and integrated on the

ﬂy depending on users’ requests. This allows a high

degree of independence and autonomy for the local

atlases and facilitates extendibility in an architecture

where atlases can be added and removed at any time

without affecting the global atlas, provided that they

implement core services including OGC CSW for

the delivery of metadata, Web Map Services (WMS)

(de la Beaujardiere, 2004) for the delivery of maps,

and WFS (Vretanos, 2005) for the delivery of vector

data.

In this article, we focus on the data discovery and

search aspects. We propose an ontology-based me-

diation approach for OGC CSWs. The solution dif-

fers from the classical mediation approaches cited in

subsection 3.2 in the way that it deals with metadata

which are already in the same format, XML, and have

the same ISO-19139 schema. It does not try to com-

bine information from different sources (i.e. atlases or

CSWs); rather it focuses on the semantic values con-

tained within the metadata records and tries to solve

semantic conﬂicts between different applications, do-

mains, organizations, or simply CSWs. Our ap-

proach is ontology-based, i.e. it uses ontologies for

representing the semantics of data values as well as

for matching concepts of local ontologies with con-

cepts of the global ontology.

4.1 Architecture

The global atlas introduced above offers a virtual

CSW, called global CSW, which acts as a CSW me-

diator and which offers uniﬁed and transparent access

to the atlases’ CSWs. As illustrated in Figure 2, users

of the global CSW are provided with a global ontol-

ogy of terms

. The user refers to the global ontology

and formulates a CSW

GetRecords

request (c.f. sub-

section 3.1) using an area of interest and keywords de-

ﬁned in the global ontology. The global CSW rewrites

the user’s request into CSW requests over the local

atlases’ CSWs using their local ontology terms, exe-

cutes the so-obtained requests, and collects metadata

records (responses) from local CWAs.

This architecture facilitates extensibility as new

catalogue services can be added and removed at any

to refer to the integrated data schema in a mediated approach as

opposed to local schemas.

The current use case topic that the ICAN group is fo-

cussing on is coastal erosion

ICSOFT 2008 - International Conference on Software and Data Technologies

194

OCA OntologyMIDA Ontology

Mapping Ontologies

Legend

CSW Query

CSW Response

Import Link

CSW

Request

CSW

Response

Global Ontology

CSW Mediator

X Ontology

MIDA

Metadata

OCA

Metadata

MIDA CSW OCA CSW X CSW

Figure 2: Ontology-Based CSW Mediation Architecture.

time without affecting the global CSW, provided that

they come with the ontologies for the terms used by

their metadata records and that mappings between

these terms and the global ontology’s terms are pro-

vided. Another advantage of this architecture is that

the global CSW acts itself as a catalogue service,

which in its turn can be queried by another external

application or even integrated in a similar CSW me-

diation architecture, as a local CSW.

4.2 Global and Local Ontologies

A (global or local) CSW uses an ontology which de-

ﬁnes the terms used as values in its metadata records

(for example thematic keywords, places, etc.). In the

initial CSW mediator prototype, we only focus on val-

ues for keywords provided as part of metadata. Con-

forming to the ISO-19115 standard, ﬁve types of key-

words are deﬁned: discipline, theme, place, temporal,

and stratum. For each atlas, an ontology of terms re-

lated to these ﬁve keyword types is deﬁned. Relation-

ships between the terms contained within one ontol-

ogy are provided as part of the same ontology.

within

Continent

County

Country

Europe

Ireland

Cork

Place

is a

"IE"

"Ireland"

name

code

within

Figure 3: Place terms from the MIDA ontology.

Figure 3 illustrates an extract of place keywords

from the MIDA ontology. Examples of places are

respectively Europe, Ireland and Cork. Relationships

between places such as ”Cork is within Ireland”, and

”Ireland is within Europe”, can be expressed. This

helps improve keyword search using an inference en-

gine. For example, if the place keywords for a dataset

only contain the term ”Cork”, and a user queries the

metadata catalogue using the place term ”Europe”

or ”Ireland” they still will get this dataset in the re-

sponse, because Cork is in Ireland and Ireland is in

Europe, which also infers that Cork is in Europe.

Ontologies are expressed in the OWL-DL lan-

guage (Herman, 2007) in the ICAN CSW mediator

prototype.

4.3 Ontology Mappings

Ontology mappings link the global ontology to the lo-

cal ontologies. This link is crucial as it is the only

means to allow the CSW mediator rewrite user re-

quests expressed with terms from the global ontol-

ogy into requests over the local CSWs using their own

terms. Therefore, they act, as semantic translators.

For each local ontology, an OWL ontology called

mapping ontology deﬁnes the mappings between the

local ontology and the global one. A mapping ontol-

ogy imports both a local and the global ontology and

deﬁnes relationships between their concepts. An ex-

ample showing extracts of the MIDA and the OCA

mapping ontologies is illustrated in Figure 4.

In Figure 4, terms preceded by preﬁx ”global”

are from the global ontology. Those preceded by

preﬁxes ”mida” and ”oca” are respectively from the

MIDA and OCA ontologies. Relationships repre-

sented with thin lines are deﬁned as part of the lo-

cal ontologies. Those represented with thick lines are

deﬁned as part of the mapping ontologies. For ex-

ample, Coastal Protection and Shore Stabilization are

deﬁned in the MIDA and OCA mapping ontologies

as narrower terms than Human Responses to Coastal

Change.

4.4 Query Rewriting and Execution

Query rewriting is the most important task of the

mediator as it refers to the process of rewriting a

user query posed over a global schema into queries

over local data sources (in this context CSWs). We

mainly focus on the

GetRecords

CSW requests in

this section. Consider a human or software user

that poses a

GetRecords

request, searching for

metadata records available through the global CSW.

For instance, the following is a

GetRecords

request

formulated by a user interested in metadata records

about data covering any region all over the world and

related to Human Responses to Coastal Change.

ONTOLOGY-BASED MEDIATION OF OGC CATALOGUE SERVICE FOR THE WEB - A Virtual Solution for

Integrating Coastal Web Atlases

195

mida:Theme

mida:CoastalProtection

mida:CoastalDefenceStructure

global:Theme

global:AgentsOfCoastalChange

global:HumainResponsesToCoastalChange

global:CoastalChangeTopic

oca:Theme

oca:ShoreStabilization

oca:Jettiesoca:Seawalls oca:RipRap

global:Theme

global:AgentsOfCoastalChange

global:HumainResponsesToCoastalChange

global:CoastalChangeTopic

Figure 4: Extracts of the MIDA and the OCA Mapping Ontologies.

http://ican.ucc.ie/srv/en/csw?

request=GetRecords&service=CSW&version=2.0.1

...

&constraint=

<?xml version="1.0" encoding="UTF-8"?>

<Filter xmlns="http://www.opengis.net/ogc"

xmlns:gml="http://www.opengis.net/gml"

xmlns:csw="http://www.opengis.net/cat/csw/2.0.2">

<And>

<PropertyIsLike wildCard="%"

singleChar="_" escape="\">

<PropertyName>keyword</PropertyName>

HumanResponsesToCoastalChange%

</Literal>

</PropertyIsLike>

<BBOX>

/csw:Record/ows:BoundingBox

</PropertyName>

<gml:Envelope

srsName=

"http://www.opengis.net/gml/srs/epsg.xml#4326">

<gml:lowerCorner>-180 -90</gml:lowerCorner>

<gml:upperCorner>180 90</gml:upperCorner>

</gml:Envelope>

</BBOX>

</And>

</Filter>

As we only deal with keywords semantics in this

paper,the query rewriting process is quite simple. The

process consists in translating the global keywords

contained in the user’s queries into local keywords

and rewriting the initial request using the so-obtained

terms. In order to do so, the mediator starts by pars-

ing the user’s request. It identiﬁes the clauses related

to keywords, and extracts the corresponding keyword

literals. For instance, in the example above, the only

clause related to keywords is the one delimited by the

ﬁrst

tag, and the corresponding

keyword is Human Responses to Coastal Change.

For each local atlas, the CSW mediator uses its in-

ference engine to obtain all the local atlas’ terms that

are equivalent to, or narrower than, the keyword lit-

eral considered. Next, the initial clause is replaced

by a disjunction of clauses, each containing a key-

word literal corresponding to one of the so-obtained

local keywords. For instance, in the example above,

the CSW mediator will translate keyword Human Re-

sponses to Coastal Change into the MIDA keywords

Coastal Protection and Coastal Defence Structure.

Thus the corresponding clause will be rewritten ac-

cording to MIDA as follows:

<Or>

<PropertyIsLike wildCard="%" singleChar="_"

escape="\">

<PropertyName>keyword</PropertyName>

CoastalProtection%

</Literal>

</PropertyIsLike>

<PropertyIsLike wildCard="%" singleChar="_"

escape="\">

<PropertyName>keyword</PropertyName>

CoastalDefenceStructure%

</Literal>

</PropertyIsLike>

</Or>

This process is repeated for each clause in the re-

quest’s ﬁlter, of course by avoiding repetition of key-

words in a disjunction of clauses. Each so-obtained

ﬁnal

GetRecords

request is sent to the corresponding

local CSW. Records obtained as results from the local

CSWs are then collected and sent back to the user.

5 IMPLEMENTATION

A ﬁrst version of the global coastal atlas prototype

has been implemented in Java, using the Jena 2

framework for inference purposes and is available at

(http://ican.ucc.ie).

The prototype allows the user to:

• Select keywords from the global ontology;

• Select an area of interest;

• Submit a query, which will generate a CSW

GetRecords

requests to the global CSW.

ICSOFT 2008 - International Conference on Software and Data Technologies

196

The CSW mediator will consult a registry of atlases

and identify the atlases that may have data within the

bounding box selected, as a bounding box is associ-

ated with each atlas representing its geographic ex-

tent in order to optimize query execution by avoid-

ing rewriting queries over CSWs with no data cover-

ing the area of interest. Next, the atlas mediator will

translate the user’s request according to the process

described in subsection 4.4, collect the responses and

send them back to the user through the graphical in-

terface.

6 CONCLUSIONS AND FUTURE

WORK

The atlas mediator prototype described in this pa-

per is a ﬁrst step towards atlas integration as part of

a new International Coastal Atlas Network (ICAN).

The prototype showcases how ontologies and ontol-

ogy mappings can be used to integrate different het-

erogeneous and autonomous atlases (or information

systems), particularly OGC CSWs.

The next step of the ICAN initiative is to integrate

WFS and CSW mediation techniques in order to de-

ﬁne a more complete approach for integrating both

data and metadata. Also, thematic information will be

considered and interfaces will be speciﬁed for sharing

this type of information, which is highly important in

atlases. The aim is to deﬁne a complete solution for

integrating CWAs.

An initial evaluation revealed that the number of

inferred keywords in rewritten queries can be dra-

matic depending on how general the user’s selected

keywords are. In some cases, one global keyword can

correspond to more than sixty local keywords of a lo-

cal atlas. This results in huge queries whose execu-

tion can be time consuming, especially when a large

number of users are connected to the global atlas at

the same time. In addition, no effort has been made to

rank metadata records according to relevance, date, or

spatial proximity. Future work will take these prob-

lems into consideration in order to optimize query

rewriting and execution as well as results presenta-

tion.

REFERENCES

Bermudez, L. E. (2004). Ontomet: Ontology Metadata

Framework. PhD thesis, Drexel University, Philadel-

phia, USA.

de la Beaujardiere, J. (2004). OGC Web Map Service In-

terface (Version 1.3.0). Open Geospatial Consortium

Inc.

Garcia-Molina, H., Papakonstantinou, Y., Quass, D., Ra-

jaman, A., Sagir, Y., Ullman, J., Vassalos, V., and

Widom, J. (1997). The TSIMMIS approach to me-

diation: Data models and Languages. Journal of In-

telligent Information Systems.

Goasdoue, F., Lattes, V., and Rousset, M.-C. (2000). The

Use of CARIN Language and Algorithms for Infor-

mation Integration: The PICSEL System. Interna-

tional Journal of Cooperative Information Systems,

9(4):383–401.

Herman, I. (2007). Semantic Web – Web Ontology Lan-

guage (OWL). W3C.

ISO (2006). ISO/PRF TS 19139 – Geographic information

– Metadata – XML schema implementation. Interna-

tional Standardization Organization.

Kashyap, V. and Sheth, A. (2000). Information Broker-

ing Across Heterogeneous Digital Data – A Metadata-

based Approach. Kluwer Academic Publishers, Nor-

well, Massachusetts, USA.

Kirk, T., Levy, A. Y., Sagiv, Y., and Srivastava, D. (1995).

The Information Manifold. In Knoblock, C. and Levy,

A., editors, Information Gathering from Heteroge-

neous, Distributed Environments, Stanford University,

Stanford, California.

Manolescu, I., Florescu, D., and Kossmann, D. (2001).

Answering XML Queries over Heterogeneous Data

Sources. In Proceedings of VLDB, pages 241–250.

Nebert, D. and Whiteside, A. (2005). OGC Catalogue Ser-

vices Speciﬁcation. Open Geospatial Consortium Inc.

Vretanos, P. A. (2005). Web Feature Service Implementa-

tion Speciﬁcation. Open Geospatial Consortium Inc.

Wiederhold, G. (1992). Mediators in the Architecture of

Future Information Systems. IEEE Computer, pages

38–49.

Wright, D., Watson, S., Bermudez, L., Cummins, V.,

Dwyer, N., O’Dea, L., Nyerges, T., Benoit, G.,

Berman, M., Helly, J., and Uhel, R. (2007). Report

on Coastal Mapping and Informatics Trans-Atlantic

Workshop 2: Coastal Atlas Interoperability. Internal,

unpublished workshop proceedings.

ONTOLOGY-BASED MEDIATION OF OGC CATALOGUE SERVICE FOR THE WEB - A Virtual Solution for

Integrating Coastal Web Atlases

197