STUDY OF DIFFERENT APPROACHES TO THE INTEGRATION

OF SPATIAL XML WEB RESOURCES

José E. Córcoles, Pascual González

Departamento de Informática Universidad de Castilla-La Mancha Campus Universitario s/n.02071.Albacete. Spain.

Keywords: Semantic Web, Integration, Spatial XML

Abstract: The research community has begun to investigate foundations for the nex

t stage of the Web, called Semantic

Web. Current efforts include the Extensible Markup Language XML, the Resource description Framework,

Topic Maps and the DARPA Agent Markup Language DAML+OIL. A rich domain that requires special

attention is the Geospatial Semantic Web. However, in order to approach the Geospatial Semantic Web, it is

necessary to solve the problem of developing an integration system for querying spatial resources stored in

different sources. In this paper, we study two different approaches to integrating spatial and non-spatial

information represented in the Geographical Markup Language (GML). The approaches studied follow

LAV (Local as View) integration. With this study we obtain the best approach to developing a real system

for querying GML resources stored in different sources.

1 INTRODUCTION

A domain that requires special attention is the

Geospatial Semantic Web (Egenhofer et al. 2002).

The enormous variety of encoding of geospatial

semantics makes it particularly challenging to

process requests for geospatial information. Work

led by the OpenGIS Consortium (OpenGIS, 1999)

addressed some basic issues, primarily related to the

geometry of geospatial features.

In order to approach the Semantic Geospatial Web,

is necessary to solve the problem of developing an

integration system for querying spatial resources

stored in difference sources. The user should view a

virtual spatial data repository in a given domain

without knowledge of the source in which each item

of data is located.

Tackling the integration of spatial information on the

Web i

not a simple task since, for example, the

sources may store large amounts of incomplete

spatial data, which may make it necessary to join the

results of queries with spatial joins. Thus, it is

necessary to apply efficient query processing

strategies that allow spatial joins to be made on the

Web (Shahabi et al. 2003). Therefore, the

application of spatial operators means a different

treatment with respect to the integration of XML

resources with only alphanumeric (non-spatial) data.

The main aim of this paper is: (1) to study two

arch

itectures

for integrating Spatial XML resources

(GML) obtained by modifying two existing

approaches, and (2) to compare each one, with the

aim of obtaining the best approach for developing a

real system for querying GML resources stored in

different sources. In our study the spatial

information is represented in the sources by GML

because it is an XML encoding for the transport and

storage of spatial/geographic information, including

both spatial features and non-spatial features. The

mechanisms and syntax that GML uses to encode

spatial information in XML are defined in the

specification of OpenGIS (OpenGIS, 2003). Thus,

GML allows a more homogeneous and flexible

representation of the spatial information.

Query mediation has been extensively studied in the

literatu

re for

different kinds of mediation models

and for the capabilities of various sources: in the

field of non-spatial integration there are several

approaches such as Tsimmis (Papakonstantinou et al.

1995), Information Manifold (Levy et al. 1996).

More directly concerned with the integration of

364

E. Córcoles J. and González P. (2004).

STUDY OF DIFFERENT APPROACHES TO THE INTEGRATION OF SPATIAL XML WEB RESOURCES.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 364-369

DOI: 10.5220/0002613303640369

 SciTePress

XML resources, it is worth noting C-Web Portal

(Amann et al. 2001) and (Amann et al. 2002). C-

Web Portal supports the integration of non-spatial

resources on the Web, and C-Web provides the

infrastructure for (1) publishing information sources

and (2) formulating structured queries by taking into

consideration the conceptual representation of a

specific domain in the form of an ontology. On the

other hand, (Amann et al. 2002) proposes a mediator

architecture for the querying and integration of Web-

accessible XML data resources (non spatial data). Its

contribution is the definition of a simple but

expressive mapping language, following a local as

view approach and describing XML resources as

local views of some global schema.

In relation to spatial data integration, there are

approaches developed by (Gupta et al. 1999) and

(Boucelma et al. 2002). (Gupta et al. 1999) extends

the MIX wrapper-mediator architecture for

integrating information from spatial information

systems and searchable databases of geo-referenced

imagery. (Boucelma et al. 2002) presents a

mediation system that addresses the integration of

GIS data tools, following a GAV(Global as View)

approach.

In order to design approaches for querying spatial

GML resources, we have based our work on two

existing studies: (Amann et al. 2002) and (Amann et

al. 2001) mentioned above. We have selected the

first approach ((Amann et al. 2002)) because it is

focused on integrating XML resources, and it can be

extended in a simple way to query GML resources

with spatial operators. The second approach has

been selected because it is an interesting approach

that makes it possible to query different resources on

the Web. By this we mean that modifying it adds the

possibility of querying GML resources. In addition,

both approaches follow a LAV (Local as View)

integration. The LAV approach facilitates the

maintenance of the integrated schema and

mediation, although query evaluation is far more

complex than the global-as-view approach where the

integrated schema is defined in terms of source

schemas. The LAV approach is therefore favoured

in the context of the integration of resources that

change significantly over time, such as Web

resources. In short, by modifying these contrasted

approaches we exploit the solution to query XML

data on the Web.

The overview of both architectures and the

modification applied to query spatial information are

shown in Section 2 and Section 3. In Section 4 we

conclude with a comparison of the two approaches,

emphasising the most important advantages and

disadvantages, in order to obtain the best approach

for developing a real system for querying GML

resources stored in different sources.

2 APPROACH BASED ON RDF

A Community Web Portal(Karvounarakis et al.

2000)(C-Web) essentially provides the means to

select, classify and access, in a semantically

meaningful and ubiquitous way, various information

resources (sites, documents, data) for diverse target

audiences (corporate, inter-enterprise, …). The core

Portal component is a Catalog holding descriptions,

i.e. metadata, of the resources available to the

community members. In order to effectively

disseminate community knowledge, Portal Catalog

organises and gathers information in a multitude of

ways, which are far more flexible and complex than

those provided by standard (relational or object)

databases. It uses the Resource Description

Framework (RDF) standard (Brickley et al. 2000)

proposed by W3C, designed to facilitate the creation

and exchange of resource descriptions between

Community Webs. In order to query the Catalog, a

query language, called RQL, is presented in

(Alexaki et al. 2001) which allows semistructured

RDF descriptions to be queried using taxonomies of

node and edge labels defined in the RDF schema.

In order to integrate the spatial information of

several spatial XML documents (GML), we have

based our work on the Community Web Portal

concept (Amann et al. 2001) with RDF and RQL, a

declarative language for querying both RDF

descriptions and related schemas. To perform this

integration, it is necessary to make some

modifications to the original approach because in the

original approach the Catalog is considered as a

collection of resources identified by URIs and it is

described using properties. However, it does not

need to use operator over the resources, only over

the properties.

GML documents (or part of) are a resource. Unlike

the original approach, it is possible to apply spatial

operators (comparatives: cross, overlap, touch;

analysis: Area, Length) over the resources provided

they represent geometry information with GML. In

order to take advantage of this fact, we have

designed two modifications with respect to the

original approach:

Extension of RQL to support spatial operators over

the resources that represent spatial documents or part

of spatial documents. These operators must be the

same as those defined in (Corcoles et al. 2001) for a

query language over GML. There are two types of

operators: methods for testing Spatial Relations and

methods that support Spatial Analysis. This

extension is not dealt with in this paper.

Extension of the Community Web Portal architecture

to support the application of the spatial operators

STUDY OF DIFFERENT APPROACHES TO THE INTEGRATION OF SPATIAL XML WEB RESOURCES

365

Figure 1: Portion of Catalog of a CityModel.

over the resources involved in the query, and the

integration of all information to be returned.

2.1 An Overview

In this section, an overview of the application of the

mediation system is given, looking at the system

from the point of view of the user. In Figure 1, an

example of the Portal Schema and its instances is

shown. The example has been obtained from the

specification documents of GML. Due to RDF’s

capability for adding new feature and geometry

types in a clear and formal manner, this example has

been carried out extending the geospatial ontology

defined by OpenGIS, where the class (Geometry,

LineString, etc) and properties (coordinates,

PolygonMember, etc) are defined. The example

shows an extension of a Catalog for a CityModel

proposed by (OpenGIS, 2003) and called

Cambridge.rdfs.

This is a well-known example used in all

specifications over GML and developed by

OpenGIS. The Cambridge example has a single

feature collection of type 'CityModel' and contains

two features using a containment relationship called

'modelMember'. The model member can be Rivers

that run through the City or Roads belonging to the

City.

An example of a query expressed in RQL extended

with spatial operators may be as follows:

“Find all Road resources that belong to Albacete

City and which are within 50 meters of a River of

this city ”.

In RQL:

Select Z

From {X}ModelMember{Y:River},

{W}ModelMember{Z:Road}, {P}name{Q}

Where X=W and X=P and Q=”Albacete” and

Crosses(Buffer(Y,50),Z)

In a different way to the original approach, this

query has a part that is executed directly over the

Catalog, and another part that is executed over the

objects Road and River stored in the respective

sources. In order to do this, it is necessary to

establish a spatial query plan. On the other hand, if

the query uses operators like Area, Length, Union,

Intersection, etc., this approach manages the new

created resources. The results could be the resources

(&r2 ,&r3).

Although the resources may be of different types

(documents, HTML files, Raster image,..), in this

approach the semantic of the spatial operators is

only applied over the geometry objects based on the

OpenGIS specification (OpenGIS, 1999).

This approach can be studied at length in [CG03]).

3 APPROACH BASED ON

MAPPING

In this section we present a general overview of the

second approach. This work has been inspired by

(Amann et al. 2002), which proposes a mediator

architecture for the querying and integration of Web-

accessible XML data resources (non spatial data). Its

contribution is the definition of a simple but

expressive mapping language, following a local as

view approach and describing XML resources as

ICEIS 2004 - DATABASES AND INFORMATION SYSTEMS INTEGRATION

366

local views of some global schema. This approach

offers its users a virtual data repository in a given

domain. This repository is virtual because the real

data resides in some external sources. However, the

users of the repository are not concerned with the

source location and source data organisation.

Obviously, our aims (integrate spatial sources) are

different from those of the previous work by

(Amann et al. 2002) (integrate non-spatial sources).

For this reason, (Amann et al. 2002) has been

extended in order to satisfy our requirements.

We present in the following section an overview of

the system architecture offered by (Amann et al.

2002). In subsection 3.2, we detail the main

modification of this approach to achieve spatial

queries.

3.1 Overview

The main task of an integration mediator is to

provide users with a unique interface for querying

the data, independently of its actual organisation and

location. This interface, or global schema, is

described as an ontology. As used here, an ontology

denotes a light-weight conceptual model and not a

hierarchy of terms or a hierarchy of concepts (in the

same way as in the first approach). The global

schema can be viewed as a simple object-oriented

data model. Hence, a global schema can be viewed

as defining a database of objects, connected by roles,

with the concept extents related by subset

relationships as per the isA links in the schema.

Since it is an integration schema, this is a virtual

database. The actual materialisation exists in the

sources.

To evaluate a user query expressed in terms of the

ontology, the approach translates it into one or more

queries on the XML sources. For this purpose, we

need to establish a correspondence between each

source and the global ontology. This correspondence

is described by a mapping, which is a collection of

mapping rules (path-to-path).

The description of the global schema in terms of the

ontology allows users to formulate structured

queries, without being aware of the source specific

structure. (Amann et al. 2002) illustrates querying

with the query language defining tree queries.

Although these queries have some limitations (e.g.

joins between variable are not allowed (Amann et al.

2002)), they are sufficiently powerful to illustrate

the issues of answering queries from XML source

data (but it is not sufficient to answer spatial

queries).

In order to process each query, two query processing

cases are possible. In the first case, the solution to a

query is the union of the complete answers from

individual sources. If no complete answer can be

obtained from a source, then the source is

abandoned. This simple strategy has obvious

advantages, as it only needs a variable binding

algorithm and a simple query execution plan for

searching all sources for which there exists a full

binding. In contrast, the second case also allows for

incomplete answers from a given source. If a source

s can only partially answer a query, then the query is

decomposed into two parts, one to be fully answered

by s and the other part being sent to the other

sources. In this case, it needs a variable binding

algorithm and a query execution plan that includes

query decomposition for searching all sources for

which there exists a full/partial binding.

3.2 Modification to query spatial

resources

Our modification includes two new components: (i)

spatial system, used to make the joins between the

results of spatial queries and (ii) Spatial DBMS,

where the spatial GML documents are stored over

ORDBMS (in the same way as the first approach).

In addition, other modifications have been made to

the functionality of the existing components: (iii)

extends the features of the query languages

including spatial operators and including spatial and

non-spatial joins, and (iv) extends the query

execution plan to query spatial joins (using the

spatial system component).

The query language used for querying in the original

approach should support spatial and non-spatial

operators with a user-friendly interface. Therefore,

in our modification we use another spatial query

language for GML (Corcoles et al. 2001) as the

query language used by the users and for querying

the different sources. Thus we have simplified the

translation between the user query and the queries

executed in each source. It is an advantage with

respect to (Amann et al. 2002). (Note that in the

original approach, a query language based on OQL

is shown.) A query is a simple tree query, based on

select-from-where clauses. We assume queries

satisfy the following restrictions regarding the

original algebra.

First, over the variables in the select clause it is

possible to apply spatial operators supported in the

original algebra: methods for testing Spatial

Relations and methods that support Spatial

Analysis(Corcoles et al. 2001). Second, as is

mentioned above, in the original approach the where

clause is a conjunction of simple predicates, where a

simple predicate is of the form x θ d in which θ ∈

{=, <,>,≤,≥}and d is an atomic value. Thus, it is not

possible to express joins by equalities between

STUDY OF DIFFERENT APPROACHES TO THE INTEGRATION OF SPATIAL XML WEB RESOURCES

367

variables, i.e. by predicates of the form x

= x

. In our

case, this limitation restricts the expressive power of

the query language and of this application. For this

reason, we have incorporated the possibility of

including spatial joins in the where clause (x

is a spatial join operator defined by (Corcoles et al.

2001); cyclic joins are not allowed). This facility

makes it more difficult to evaluate the queries than

in the original approach. For this reason, the

definition of a new algorithm (shown below) for

evaluating the query is necessary. Third, Spatial

Operators intersection, union, difference and

symdifference are not included. This restricts the

power of the query language but simplifies the

evaluation of queries. Last, the language has no

quantifiers, aggregates, or subqueries.

The result of such a query is a set of tuples of the

form {[a

, a

, …a

]}where a

, a

, …a

are instances of

the variables in the query’s select clause and can be

either atomic values or GML fragments.

Due to the extension of the original approach with

spatial and non-spatial joins, a third query

processing case should be added in this

modification. This case allows for incomplete

answers from a given source and all variables

involved in spatial join operators (or non spatial)

are found in different sources. It is the most

complete and complex case. It needs (i) a query

execution plan that includes query decomposition

for searching all sources for which there exists a

full/partial binding, (ii) a solution for joining the

results of each partial query, and (iii) a strategy for

performing spatial joins on the web sources. For this

reason we have developed a new query execution

plan, which is described in the following section.

3.3 Query Execution Plan

This Section describes a query execution plan that

includes query decomposition for searching all

sources for which there exists a full/partial binding

and a strategy for performing spatial joins on the

web sources.

In a distributed environment with restricted access to

the remote servers (the remote servers are read-

only), performing spatial join queries must be

simulated by spatial select operations after

transferring whole/partial data sets from the local to

the remote server. In addition, the query response

time is a function of size and complexity of the data

transferred between the servers. (Shahabi et al.

2003) provides an approach to performing spatial

joins in a Web environment. Based on this work, we

have adopted the following query processing

strategy to solve spatial joins. Local refers to the

mediator and Remote refers to the remote sources.

(i) Local: Local to Remote Transfer {Dinamic-

MBR}, (ii) Remote: Spatial Selection {Window-

Selection}, (iii) Remote: Send to Local {Candidate

Objects} and, (iv) Local: Refinement {Pipelined}.

Given a set of sources S and a query Q, the

algorithm P(Q) shown in figure 2 computes a Query

Execution Plan for Q. For each source s and a

maximal binding β ∈ B(Q,s), a QEP P(β) is

computed: if β is a full binding (i.e complete

answers are obtained), the result is query Q.

Otherwise, if β is a partial binding, then query Q is

decomposed into a prefix query Q

(β) and a set of

suffix queries QS(β). All spatial joins divided in the

prefix query are registered. It is necessary to know

when a partial query satisfies a spatial join and then

carry out the join in the QEP. The query execution

plan of Q against source s is obtained by joining

(β) with the query execution plan for each suffix

query Q´∈ QS(β) (variable k denotes the key query

variables of Q´). To calculate the query execution

plan of a suffix query Q’, the algorithm is called

recursively. Finally, the plan obtained is added to the

existing plan by union.

Figure 2: Query Execution Plan Generation.

Note that there are two reasons for interrupting the

calculation of a query execution plan for a given

source s and binding β. The most trivial case is that

there exists no maximal binding for Q in s. The

second reason is that there exists at least one suffix

query which cannot be satisfied (empty query

execution plan).

4 CONCLUSIONS

Each approach has advantages and disadvantages.

With the modification incorporated in the first

approach it is possible to obtain an architecture that

enables the integration of different kinds of

ICEIS 2004 - DATABASES AND INFORMATION SYSTEMS INTEGRATION

368

resources (documents, sites, GML resources,...)

using a unique query language RQL that allows

querying of a Catalog with references to all

resources. In addition, this approach (and the second

approach) allows GML resources to be queried

efficiently in each source, because they store the

GML documents in ORDBMS(Corcoles et al. 2002).

These are just some of the features offered by the

first approach.

No doubt, this approach is the best approximation to

the Geospatial Semantic Web, querying all kinds of

resources in the same way. However, this alternative

has several disadvantages: (1) the query language

used by the user (RQL) is different from the query

language used by the wrappers. For this reason,

several conversions are necessary between query

languages (RQL to QL over GML and QL over

GML to QL of the DBMS); (2) the rewriting

algorithm to translate the RQL query to XML

queries over the local sources is complex to

implement.

The second modification offers a more powerful

alternative for querying spatial resources. It solves

some disadvantage of the first approach: (1) the use

of only one query language to query the ontology

(users) and to query the GML resources in the

wrappers. The approach architecture allows a more

powerful query language than the modification of

RQL; (2) different strategies are contemplated in this

approach, including spatial joins between objects

localised in resources of different sources. This

approach uses strategies for carrying out joins

efficiently between large spatial data on the Web.

Finally, both approaches also has the disadvantage

of administering the mapping rules between the

ontology and the sources.

In conclusion, an approach with advantages of both

alternatives is the more powerful alternative for

developing a system to query spatial XML resources

on the Web. The implementation of a usable

prototype has already been achieved.

REFERENCES

Amann B., et al.. 2002 Ontology-Based Integration of

XML Web Resources. In International Semantic Web

Conference (ISWC), Sardinia, Italy.

Alexaki. S et al. 2001. “The ICSFORTH RDFSuite:

Managing Voluminous RDF Description Bases”. In

Proceedings of the 2nd International Workshop on the

Semantic Web (SemWeb'01), in conjunction with

WWW10, pp. 1-13, Hong Kong.

Amann, B. et al. 2001. Mapping XML Fragments to

Community Web Ontologies. In Proc. Fourth

International Workshop on the Web and Databases.

Boucelma, O, et al. 2002. A WFS-Based Mediation

System for GIS Interoperability. ACM-GIS 2002. 10th

ACM International Symposium on Advances in

Geographic Information Systems. McLean (USA).

Brickley, D , et al. 2000. Resource Description Framework

(RDF) Schema Specification 1.0, W3C Candidate

Recommendation. Technical Report CR-rdf-schema-

20000327, W3C, Available at

http://www.w3.org/TR/rdf-schema.

Córcoles J, et al. 2001. A Specification of a Spatial Query

Language over GML. ACM-GIS 2001. 9th ACM

International Symposium on Advances in Geographic

Information Systems. Atlanta (USA).

Córcoles J, et al. 2002. Analysis of Different Approaches

for Storing GML Documents ACM-GIS 2002. 10th

ACM International Symposium on Advances in

Geographic Information Systems. McLean. (USA).

Córcoles J, et al. 2003. Querying Spatial Resources. An

Approach to the Semantic Geospatial Web. CAiSE'03

workshop (WES2003).

To Appear in Lecture Notes in

Computer Science (LNCS)

by Springer-Verlag.

Egenhofer M. 2002. Toward the Semantic Geospatial

Web. ACM-GIS 2002. 10th ACM International

Symposium on Advances in Geographic Information

Systems. McLean (USA).

OpenGIS. 2003. Geography Markup Language (GML)

v3.0. http://www.opengis.org/techno/documents/02-

023r4.pdf.

Gupta, A, et al.. 1999. Integrating GIS and Imagenery

through XML based information Mediation. Integrated

Spatial Databases: DigitalImages and GIS. Lecture

Notesin Computer Science. Vol1737. Pp. 211-234.

Springer-Verlag.

Karvounarakis, G, et al. 2000. Querying community web

portals. Technical report, Institute of Computer

Science, FORTH,Heraklion, Greece.

http://www.ics.forth.gr/proj/isst/RDF/ RQL/rql.pdf.

Levy,A. Y, 1996. Querying Heterogeneous Information

Sources Using Source Description. In Proc. of the Int.

Conference on Very Large Databases, pp.25-262.

India.

OpenGis Consortium. 1999. Specifications.

http://www.

opengis.org/ techno/specs.htm

Papakonstantinou, Y, et al. 1995. Object Exchange Across

Heterogeneous Information Sources. In Proc. ICDE

Conf. TSIMMIS project: http://www-

db.standford,edu/tsimmis.

Shahabi, c, et al. 2003. Alternative Strategies for

Performing Spatial Joins on Web Sources,

To Appear

in Journal of Knowledge and Information Systems

(KAIS) by Springer-Verlag

STUDY OF DIFFERENT APPROACHES TO THE INTEGRATION OF SPATIAL XML WEB RESOURCES

369