IMPROVING ANALYSIS PATTERNS IN THE GEOGRAPHIC

DOMAIN USING ONTOLOGICAL META-PROPERTIES

Evaldo Silva, Jugurta Oliveira and Gabriel Gonçalves

Department of Informatics, Federal University of Viçosa, 36570-000, Viçosa, Brazil

Keywords: Conceptual Modeling, Analysis Pattern, Ontology, Sharing of Knowledge.

Abstract: This article shows the improving of analysis patterns in the geographic domain through ontological meta-

properties, where each pattern’s class has its concept analyzed ontologically. This improvement also permits

to restructure the class diagram of analysis patterns, increasing the reuse quality. Besides improving the

analysis patterns, it is proposed one more topic in the template for analysis patterns documentation. This

topic is based on the specification of the main classes defined during the process.

1 INTRODUCTION

Some researchers consider ontologies and patterns

as mechanisms used during the phases of

specification and construction of information

systems (Guizzardi et al., 2002) (Devedzic, 1999)

(Hamza, 2004). The both are also defined as

mechanisms to reuse of knowledge. However,

ontologies are created by domain specialists, where

knowledge is reusable in a widely way. Analysis

patterns are discovery and reused in application

design level, by information systems designers.

According to Guarino (1995) “ontology is a

specification of a conceptualization described by

concepts and relationships that can exist in the real

world. The conceptualization can be developed

containing terminologies and vocabularies,

establishing properties and allowing the knowledge

to be reused, avoiding the rework or rediscover of

equivalent terminologies.”

The conceptual schema of an analysis pattern is

modelled through collection of classes and

associations, whose follows the notation of UML

(Unified Modeling Language), it has some meaning

in the context of an application, where the same

structure can be considered valid to other

applications. That causes the same conceptual

schema to be reused to compose the modeling of

other information systems (Fernandez, 1998).

There are researches that have been developed

aiming the improvement of the reuse of analysis

patterns through ontological approaches, increasing

the productivity and quality during the conceptual

modeling of systems (Guizzardi et al., 2002)

(Devedzic, 1999) (Hamza, 2004).

Guizzardi et al. (2002) propose an approach to

derive frameworks from domain ontologies. They

have shown how to implement an application of

information systems using such approach in the

developing process. According to the authors,

domain ontologies are used to support the analysis

activity, it is necessary its transformation in the view

of class diagram, without losing the explicit

representation of the knowledge.

Devedzic (1999) approaches that the concepts

used in ontologies and software patterns can

superpose each other, enriching the knowledge about

the domain. He proposes the use of software patterns

as a source of knowledge during the process of

conception and developing of ontologies. In a

similar way, Hamza (2004) proposes an ontological

approach to improve the quality of analysis patterns

reuse, where the knowledge taken from the patterns

is used to the development of ontologies. The

patterns are grouped and reused through ontologies,

enlarging the knowledge about the domain.

Hamza (2004) also points some problems related

to the reuse of analysis patterns, which can be

resolved through an ontological approach: (a) the

lack of formalism; (b) redundancy of patterns; (c)

indefiniteness of the reuse domain.

In this way, this article shows the improvement

of analysis patterns through ontological meta-

properties in the geographic domain. The

improvement permits restructure the class diagram

256

Silva E., Oliveira J. and Gonçalves G. (2008).

IMPROVING ANALYSIS PATTERNS IN THE GEOGRAPHIC DOMAIN USING ONTOLOGICAL META-PROPERTIES.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - ISAS, pages 256-261

DOI: 10.5220/0001697002560261

 SciTePress

of analysis patterns, increasing the quality of reuse.

It is also proposed the documentation of analysis

patterns based in the specification of the main

classes defined during the process.

To exemplify the process of analysis patterns

improvement, it will be used the Parceling of Urban

Land analysis pattern, proposed in Lisboa et al.

(2002). As a technique to the application of

ontological meta-properties, it will be used the

VERONTO (ONTOlogical VERification) developed

by Villela (2004). VERONTO technique is used to

the verification and adequacy of class diagrams of

UML (Unified Modeling Language), based on the

philosophical notions of rigidity, dependency and

identity, defined by Guarino and Welty (2000).

This article is organized as follows. Section 2

describes the VERONTO technique. Section 3

discusses the analysis patterns documentation using

one more topic, which identifies the main classes of

the domain. Section 4 shows the improvement of the

Parceling of Urban Land analysis pattern. Section 5

presents some final considerations.

2 VERONTO TECHNIQUE

The VERONTO technique, proposed by Guarino

and Welty (2000), uses meta-properties as rigidity,

dependence and identity, in the validation of

conceptual models specified through UML class

diagrams. Ontological meta-properties are based on

philosophical notions of essence, dependence and

identity defined by Guarino and Welty.

When developed this technique, Villela et al.

(2004) applied these meta-properties in elements of

the class diagram, making it possible to apply

taxonomic restrictions about the relationship

between classes. Such restrictions are based on

ontological analysis of meta-properties, applied in

the elements of the class diagram.

2.1 Representation of Philosophical

Notions through Ontological

Meta-Properties

The notion of essence is represented by the meta-

property rigidity. The meta-property of rigidity is

about the knowledge of how classes can change in

the course of time and other can’t. A meta-property

is rigid (+R) (~, anti-rigid) when an element of the

domain that instantiates such property will continue

to instantiate it during all its existence (Guarino and

Welty 2000). For example, in a conceptual schema

which specifies an application of urban transport, an

instance of the FAST TRAFFIC ROAD class can

stop being a fast track road to become a local road,

but it will always be a road within the context of the

transportation system. Thus, it is possible to analyze

that the FAST TRAFFIC ROAD class is anti-rigid

(~R), as a fast traffic road will not be like that for all

its existence. However, the ROAD class is rigid

(+R), because in the domain of urban transport

application, an instance of ROAD will be like that

for all its existence.

The philosophical notion of dependence is about

relations of dependency that can be intrinsic and

extrinsic, represented by +D (-D, otherwise)

(Guarino and Welty 2000).

According to Villela et al. (2004) dependency

involves different relationships, such as the ones

existing among people and their parents, being

extrinsic and intrinsic. An intrinsic property is inhe-

rent to the individual, non dependent on other indivi-

duals, like having a heart or a fingerprint. Extrinsic

properties are not inherent and they have a relational

nature, like “being the mayor of the São Paulo City”.

For example, an instance of the DISTRICT SEAT

class is externally dependent on the MUNICI-

PALITY class, as it can only be a district seat if

there is a municipality in which it was created.

At last, identity is about the way we recognize

individual entities, and it is based on the concept of

Identity Condition (IC), proposed by Guarino and

Welty (2000). A class that has an identity condition

is represented by the symbol +O (-O, otherwise),

only if it is rigid and executes an IC (+I) (-I,

otherwise). A non-rigid class can execute an IC, if

and only if this is inherited by a class that has a rigid

meta-property, which subsume it. For example, the

subclass FAST TRAFFIC ROAD, classified as non-

rigid, can only execute its IC’s, inheriting them from

rigid meta-properties that superpose it, as a meta-

property (+O) from the super class ROAD.

The meta-properties described above create some

natural restrictions in the taxonomic structure of

ontology (Guarino and Welty, 2000), supporting the

analysis and adequacy of conceptual models. Be it

two arbitrary classes (Φ and ψ). The notation Φ

used to show that a class Φ has the meta-property

with the restrictions showed in Table 1.

Table 1: Taxonomic restrictions (Guarino, 2000).

Meta-Properties Restrictions

Rigidity

can't subsume ψ

Identity

can't subsume ψ

-I

Dependency

can't subsume ψ

-D

IMPROVING ANALYSIS PATTERNS IN THE GEOGRAPHIC DOMAIN USING ONTOLOGICAL

META-PROPERTIES

257

Besides the meta-properties seen before, it is part of

the VERONTO technique the addition of rules to the

use of relationships, based on the rules proposed by

Wand et al. (1999) that follow:

a) Rules to optional relationships:

• Optional associations (minimum cardinality 0)

must be avoided.

• Acquisition or loss of an association must be

modeled as a change of class.

• The capacity of instances of a class to participate

in an association, without losing properties, must

be modeled as a sub classification.

b) Rules to aggregation

• Each class component must be associated with a

class composed through an aggregation.

• The emergent properties of the composed class

(“whole”) must be modeled as attributes and

associations.

Practical implications of such rules, as well as some

advantages are discussed by Villela (2004).

3 DOCUMENTING ANALYSIS

PATTERNS WITH THE RIGID

CLASSES

The analysis patterns description is important to

analysis patterns should be reused. Besides, it serves

as a repository, allowing the sharing of the

knowledge by designers of information systems.

The analysis patterns template proposed on this

work follows the model proposed in Meszaros and

Doble (1998), which defines the following topics

that specify the catalog: Name, Problem, Context,

Forces, Solution, Participants, and Related Patterns.

Nevertheless, this specification does not follow a

generic format. Some analysis patterns are

documented and cataloging in a narrative form, but

generally there are coincidences in the description of

the analysis patterns in a topic structure. Though

there is not a generic topic structure to use, the

topics listed before have a basic format to the

specification of analysis patterns.

Based on the template’s structure mentioned in

Meszaros and Doble (1998), it will be specified one

more topic, called Rigid Classes. The Rigid Classes

topic is based upon the taxonomy concept

“backbone” defined by (Guarino and Welty 2000).

The “backbone taxonomy” is defined by rigid classes

found in the hierarchical structure of ontology.

Those classes model the main concepts, embracing

the entire domain and describing the basic structure

which serves as a reference to the reuse of

knowledge.

Thus, after the improvement of patterns through

the VERONTO technique, it is also possible to

identify the rigid classes that exist in the analysis

patterns. The documentation based on the rigid

classes aims at helping the designer in the reuse of

patterns, identifying the main classes of the domain.

4 IMPROVING THE PARCELING

OF URBAN LAND PATTERN

The patterns improvement process occurs through

the ontological analysis, classifying each class based

on the meta-properties described before.

To develop this section, it was chosen the

Parceling of Urban Land analysis pattern (Figure 1),

proposed in (Lisboa F. et al. 2002). The diagram

follows the UML-GeoFrame model, a UML profile

to model geographic database (Lisboa F. and Iochpe

2008). It was depicted using a CASE tool called

ArgoCASEGEO (Lisboa F. et al., 2004).

The result of the ontological analysis about the

Parceling of Urban Land pattern is the following.

According to Demographic Census from IBGE -

Brazilian Institute of Geography and Statistics

(IBGE 2008), a city is classified as an urban area of

city (municipal seat) or village (district seat).

Therefore, a village or district seat, for example, can

be emancipated and become municipal seat. So, the

CITY class is classified as: Anti-rigid (~R) – all city

will not be like that for all its existence; it does not

give identity (-O) – all city can exist in several urban

regions as district or municipal seats; it executes

identity (+I) – it executes identity inherited from

seat; it is non-dependent (-D) – it does not depend

on any other class.

The ADMINISTRATIVE DIVISION class

models the subdivision of municipal territory. It is

classified as: Rigid (+R) – every administrative

division will be like that for all its existence; It gives

identity (+O) – all instance of the ADMINISTRA-

TIVE DIVISION class can be identified through a

territorial division; it executes identity (+I) – if it

gives identity it can also execute it; it is dependent

(+D) – it is externally dependent on a municipality.

ICEIS 2008 - International Conference on Enterprise Information Systems

258

Figure 1: Parceling of Urban Land Analysis Pattern (Lisboa Filho et al. 2002).

The DISTRICT class is defined as an administrative

unit of a municipality. It is classified as: Anti-Rigid

(~R) – a district can emancipate itself, therefore, it

will not be like that for all its existence; it does not

give identity (-O) – the existence of the

administrative division of a municipality identifies

the districts; it executes Identity (+I) – it executes an

identity inherited from the ADMINISTRATIVE

DIVISION class; it is Dependent (+D) – it is

externally dependent on a municipality.

The QUARTER class stores information about

the intra-urban regions of a city or village, and

through a municipal law, can become an administra-

tive unit (IBGE 2008). It is classified as: Anti-Rigid

(~R) – every instance of quarter it will not be like

that for all its existence; it does not give identity

(-O) – the quarter is identified from the existence of

an intra-urban region; it executes identity (+I) - it

executes identity inherited from the definition of an

intra urban region it is non-dependent (-D) – it does

not depend on any other class.

The BLOCK class stores information about

urbanized areas of a city. A block can be created

through a municipal law (IBGE, 2008). It is

classified as: Rigid (+R) – every instance of the

BLOCK class it will be like that for all its existence;

it gives identity (+O) – every block can be identified

within the urban area, through its spatial

localization; it executes identity (+I) – if it gives a

condition of identity, it also executes it; it is

dependent (+D) – it is externally dependent on a

municipality.

The PARCEL class stores information about a

portion of a parceled land, with its front to public

way and meant for receiving a construction. It is

classified as: Rigid (+R) – every instance of parcel it

will be like that for all its existence; it gives identity

(+O) – every parcel can be identified within its ur-

ban area, during its existence, through its spatial

localization; it executes identity (+I) – if it gives a

condition of identity, it also executes it; it is not de-

pendent (-D) – it does not depend on any other class.

The FOREFACE class stores information about

the alignment of a parcel or group of parcels facing

the same road. It is classified as: Rigid (+R) – all

foreface of parcel will be like that for all its

existence; it gives identity (+O) – every foreface can

be identified within the urban area, through its

spatial localization; it executes identity (+I) – if it

gives a condition of identity, it also executes it; it is

dependent (+D) – it is externally dependent on the

PARCEL class.

The SIDE OF THE BLOCK class stores

information about the composition of the Foreface of

the parcels. It is classified as: Rigid (+R) - every

instance of the SIDE OF THE BLOCK class will be

like that for all its existence; it gives identity (+O) –

every side of the block can be identified within the

urban area, through its spatial localization; it

executes identity (+I) – if it gives a condition of

identity, it also executes it; it is dependent (+D) – it

is externally dependent on the BLOCK class.

4.1 Adequacy to the Restrictions

Imposed by Meta-Properties

Following the technique VERONTO (Villela et al.

2004), the CITY class must be a subclass of a rigid

IMPROVING ANALYSIS PATTERNS IN THE GEOGRAPHIC DOMAIN USING ONTOLOGICAL

META-PROPERTIES

259

class so that it can inherit a condition of identity. In

this case, it was included the SEAT class as

superclass of CITY, with the meta-property rigid

(+R), giving an identity (+O). Following the same

principle to the QUARTER class, it was created a

superclass, called INTRA-URBAN REGION, with

the meta-properties rigid (+R), giving an identity

(+O).

The DISTRICT subclass must continue as a

subclass of the ADMINISTRATIVE DIVISION

superclass. In other words, according to IBGE

(2008) a district is considered a subdivision of the

political-administrative organization of a

municipality.

The VILLAGE class can also be suggested and

modeled in the analysis of the hierarchical

relationship. According to IBGE (2008), a village is

a District Seat, therefore it was suggested as a

subclass of the SEAT class.

In a similar way the SEAT class, it was also

included the MUNICIPALITY class, since it is

through it that it is possible to have relationships

with the municipal seat and the administrative

division.

Besides the hierarchical relationships, it is part of

adequacy of pattern the verifying of optional and

aggregation relationships through the rules proposed

by (Wand et al. 1999).

The association between MUNICIPALITY and

SEAT must be mandatory, that is, every

municipality must have a seat. The CITY and

VILLAGE classes can have one or several urban

regions, because are considered as an urban area of

city (IBGE, 2008).

The association between CITY and

ADMINISTRATIVE DIVISION comes to an end,

because it is the municipality that has administrative

units, having an association with the Administrative

Division.

The association between QUARTER, BLOCK,

PARCEL, FOREFACE and SIDE OF THE BLOCK

obey to the rules of aggregation and, therefore, will

not be altered.

Figure 2 shows the Parceling of Urban Land

analysis pattern revised according to the technique

VERONTO, and then it is shown the specification of

analysis pattern according to the topics presented in

the item 3.

The Rigid Classes topic specified in Table 2

allow identifying the main classes of the domain.

These classes can help in reuse of analysis

pattern during the conceptual modeling of

Geographic Information Systems (GIS). Besides,

through analisys pattern catalog, the Parceling of

Urban Land Pattern can be reusable for several

designers, in diferents regions of country. For

example, the catalog could be published in a SDI

(Spatial Data Infrastructure), where the analysis

pattern can be shared in a widely way.

Figure 2: Parceling of Urban Land Analysis Pattern.

ICEIS 2008 - International Conference on Enterprise Information Systems

260

Table 2: New description of the Parceling of Urban Land analysis pattern.

Problem

How to structure a database of urban cadastre?

Context

The aim is to permit the manipulation of geographic information to support the public

administrator. Those databases are used by many applications as distribution of health stations,

taxes charging, and others. This kind of application needs information about the geographic regions

within a municipality.

Forces

It is not financially easy to create a base with the aimed scale. The bigger the scale the bigger the

cost to the creation and maintenance of data. A lot of kinds of regions are used depending on the

size and organization of the city. The most common kinds include quarters, administrative divisions

or urban zones.

Solution

Parceling_of_Urban_Land.xmi (Figure 2)

Participants

The Municipality class has mandatory a seat, or urban area of city, and can have administrative

units. An administrative unit can be a district or sub district. The quarter class is related to the

Block class, through a multiplicity of one-to-many. But this association must be adapted to each

specific situation, according to the municipal law.

Related Pattern

Circulation network

Rigid Classes

Municipality, Administrative Division, Municipal Seat, District Seat, Intra-urban Region, Parcel,

Foreface, Block, Side of the Block

5 FINAL CONSIDERATIONS

The process of improvement of analysis patterns

shown in this article permits to restructure the

pattern’s class diagram from an ontological analysis.

The pattern catalog according to its domain and rigid

classes, it is esteemed that it will be possible to reuse

it more widely with a well-defined purpose.

With the improvement process and the analysis

patterns catalog through ontological meta-properties,

it is possible to integrate them with domain

ontologies. Through this integration, it is possible to

create mechanisms to recover analysis patterns

through the existing knowledge in the ontology,

increasing the quality and productivity of conceptual

modeling of information systems.

ACKNOWLEDGEMENTS

This project was partially supported by the Brazilian

National Research Council (CNPq), MCT/CTINFO

and Fapemig.

REFERENCES

Devedzic, V. (1999) Ontologies: Borrowing from

Software Patterns. Intelligence.

Fernandez, E. B. (1998) Building systems using analysis

patterns. Procs. 3rd Int. Soft. Architecture Workshop

(ISAW3), Orlando, FL, November 1998. p. 37-40.

Fowler, M. (1997) Analysis Patterns: Reusable Object

Models. Menlo Park, CA: Addison Wesley Longman.

Guarino, N (1995) Formal Ontology, Conceptual Analysis

and Knowledge and Representation. Internacional

Journal of Human and Computer Studies, 43(5/6).

Guarino, N., Welty, C. (2000) A formal ontology of

properties. In R. Dieng, Ed., Proc. 12th Int. Conf. on

Knowledge Engineering and Knowledge Management,

Springer Verlag.

Guizzardi, G., Falbo R.A., Pereira Filho, J.G. (2002)

Using Objects and Patterns to Implement Domain

Ontologies, Journal of the Brazilian Computer

Society, vol. 1, no. 8.

Hamza, H (2004) Improving analysis patterns reuse: an

ontological approach. Proc. of Ontologies as Software

Engineering Artifacts Workshop, OOPSLA'04.

IBGE-Instituto Brasileiro de Geografia e Estatística (2008)

Censo Demográfico. Available in: http://www.ibge.

gov. br. Access in Jan, 29th 2008. (in Portuguese)

Lisboa Filho, J., Iochpe, C., Borges, K. A. V. (2002)

Analysis patterns for GIS data schema reuse on urban

management applications. CLEI Electronic Journal,

v.5, n.2.

Lisboa Filho, J., Sodré, V.F., Daltio, J., Rodrigues, M.F.,

Vilela, V. (2004) A CASE tool for geographic

database design supporting analysis patterns. Proc.

Conceptual Modeling for GIS, ER2004. LNCS 3289.

Lisboa Filho, J., Iochpe, C. (2008) Modeling with a UML

profile. In: Shashi Shekhar and Hui Xiong.

Encyclopedia of GIS. Germany: Springer-Verlag.

Meszaros, G.; Doble, J. (1998) A pattern language for

pattern writing.

Villela, M. L. B., Oliveira, A. P., Braga, J. L. (2004)

Modelagem ontológica no apoio à modelagem

conceitual. XVIII Simpósio Brasileiro de Engenharia

de Software. Brasília. (in Portuguese)

Wand, Y, Storey, V. C., Weber, R. (1999) An ontological

analysis of the relationship construct in modeling

conceptual. ACM Transactions on Database Systems,

24(4): 494-528.

IMPROVING ANALYSIS PATTERNS IN THE GEOGRAPHIC DOMAIN USING ONTOLOGICAL

META-PROPERTIES

261