Building Application Ontologies through Knowledge System Goals

Luis Eduardo Santos and Rosario Girardi

Federal University of Maranhão, Computer Science Departament, São Luiz, Maranhão, Brazil

Keywords: Application Ontologies, Intestate Succession, Knowledge-based Systems, Knowledge Bases.

Abstract: This article presents a case study in building an application ontology for the development of a knowledge-

based system in the field of Inheritance Law, developed with GAODT, a goal oriented technique. GAODT

proposes the translation of the goals necessary to build a knowledge-based system expressed in natural

language to rules in First-order logic, from which the elements that constitute the application ontology

(classes, relations, properties and axioms) are extracted. It is also presented a comparative evaluation

between GAODT and other state of the art techniques.

1 INTRODUCTION

Ontologies are knowledge representation structures

capable of expressing a set of entities in a given

domain, their relationships and axioms, being used

by modern knowledge-based systems (KBS) as

knowledge bases to represent and share knowledge

of a particular application domain. They allow

semantic processing of information and a more

precise interpretation of data, providing greater

effectiveness and usability than traditional

information systems (Girardi, 2010).

An ontology is classified according to its

generality, as high-level, domain, task or application

ontology (Guarino, 1998). High-level ontologies

describe generic concepts like time and space,

independently of a particular domain. Domain

ontologies make explicit concepts of a domain and

their relationships, for example, the concepts

“client”, “legal-case” are the relationship “has(client,

legal-case)” in the legal field. Task ontologies

describe the activities of a domain, for instance,

similarity analysis in the information retrieval

related activities. Finally, application ontologies are

specializations of domain and task ontologies, being

used in a particular application, for example, the task

relationship “similarity analysis” between the

concepts “old legal case” and “new legal case” in a

legal information retrieval system. According to

Guarino (Guarino, 1998), this hierarchy promotes

the reuse of ontologies, i.e., to build application

ontologies it is necessary to extend both domain and

task ontologies, and these in turn, extend high-level

ontologies. However, in practice, building reusable

ontologies is a costly process. Therefore, building

application ontologies first and then generalizing

them to domain and task ontologies is a suitable

alternative (Girardi, 2010).

Several techniques have been developed to

support the process of ontology construction.

However, most of them focus just on the

development of domain and task ontologies.

Appropriate techniques for the development of

application ontologies are needed and the GAODT

(“Goal-Oriented Application Ontology Development

Technique”) technique described in this paper

contributes to this goal.

GAODT translates the goals in language natural

expressing the requirement of a KBS to rules and

facts in First-order logic (FOL) (Russell and Norvig,

2004) and then extracts the elements that constitute

the application ontology.

This article describes a case study in building an

application ontology using GAODT to support

decision making in Intestate Succession, a type of

succession disciplined by Inheritance Law. Intestate

Succession comprises the set of rules governing the

transfer of assets to someone after his death

according to the law (Gonçalves, 2009).

The paper is organized as follows. Section 2

presents the case study, emphasizing the advantages

of the goal-oriented development cycle adopted by

the GAODT technique. Section 3 discusses a

comparative evaluation between GAODT and some

representative state of the art techniques. Section 4

concludes the paper and points out future work.

115

Santos L. and Girardi R..

Building Application Ontologies through Knowledge System Goals.

DOI: 10.5220/0004143501150124

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2012), pages 115-124

ISBN: 978-989-8565-30-3

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

2 A CASE STUDY ABOUT

INTESTATE SUCCESSION

The main concepts about Intestate Succession

considered in this work are following described.

Intestate Succession is a legal institute that

governs the transfer of property of a person by the

reason of his/her death without a will. In that case

there is set of rules used to determine who will

inherit the property.

In the Brazilian law, the following order is

applied: First, the descendants (children,

grandchildren, and so on) concurring with the

spouse; in the absence of descendants, the

ascendants (father, mother, grandfather,

grandmother, etc.) also concurring with the spouse;

if there aren’t ascendants, the spouse and finally, in

the absence of a spouse, the properties are

transmitted to the collaterals (article 1829 of

Brazilian Civil Code).

The concurrency of the spouse with descendants

or ascendants depends on the matrimonial regime.

Matrimonial regime, are systems of property

ownership between spouses providing for the

creation or absence of a marital estate, and if

created, what properties are included in that estate,

how and by whom it is managed, and how it will be

divided and inherited at the end of the marriage,

which can be of four types: Universal Community

Regime, Limited Community Regime, Accrual

Regime and Separation of Property.

The Universal Community Regime is the union

of all pre-marital and marital property of a couple,

and when sharing, the surviving spouse does not

compete for the inheritance, since half (moiety) of

all properties belong to him/her. For example,

consider that the late “John” left properties valued at

R$ 100,000. Half of this amount belongs to “Mary”,

his alive wife, and the remaining money will be

divided among the other heirs.

In Limited Community Regime, the surviving

spouse owns half of the marital properties and still

competes for the inheritance of pre-marital

properties with the descendants. For example, the

late “Peter” left pre-marital properties equivalent to

R$ 200,000 and a patrimony made with “Louise”

valued at R$ 100,000. The spouse already owns half

of the marital properties and will also compete with

the descendants for pre-marital properties.

In the present case study only these two regimes

will be considered.

2.1 An Overview of the GAODT

Technique

Figure 1 illustrates the GAODT process along with

its four activities: “Selection of Goals and Facts”,

“Representation of Predicates in FOL”,

“Specification of Axioms in FOL” and

“Specification/Extension of the Application

Ontology”.

The developer of the application ontology and

the domain expert participate in the execution of the

activities. The developer is the knowledge engineer

responsible for building the application ontology.

The domain expert is someone who has expertise in

an area of knowledge.

The technique takes as input a list of all the goals

and facts of the system provided by the domain

expert. The goals are the requirements that the KBS

has to achieve, for instance, “Calculate the

inheritance of a person” and the facts are general

statements like “A person might have descendants”.

In the activity “Selection of Goals and Facts”, the

developer, in consensus with the domain expert,

selects the most representative goals and facts to be

used as input of next activity. In the activity

“Representation of Predicates in FOL”, the

developer translates the goals and facts in natural

language to predicates in FOL.

The activity “Specification of Axioms in FOL”

takes as input the predicates specified in the

previous activity and specifies in FOL the rules

needed to achieve the goals of the system. This

activity is iterative, that is, a goal predicate may

require the achievement of other subgoals. For

example, to satisfy the goal “Determine the

ascendants of a person”, other subgoals should be

achieved, such as “Determine the genitor of a

person”. All the process is iteratively executed until

all the goals have been decomposed and expressed

as simple facts. Finally, the activity

“Specification/Extension of the Application

Ontology” uses axioms generated on the previous

activity and extracts from them the necessary

elements to compose the application ontology. The

created application ontology can be extended by

performing a semantic search in a repository of

application ontologies. In the next subsections

GAODT activities are explained in further detail.

2.2 Selection of Goals and Facts

This activity takes as input a list of all the goals and

facts of the system, provided by the domain expert.

From this list, the developer and the specialist sets

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

116

Figure 1: An overview of the GAODT technique.

which of them will be given as input to the next

activity. As an example, Table 1 presents a list of

goals and facts for an application ontology for a

knowledge-based decision support system on

Intestate Succession.

Initially it must be defined the general goal and

the main specific goals of the system. For instance,

for the general goal 1: “Calculate the inheritance of a

person”, the main specific goals are 2: “Identify the

heirs of a person” and 3: “Determine the inheritance

of each heir”. To satisfy these subgoals, other goals

could be defined in subsequent iterations, in a

process performed recursively for all the goals in the

list.

2.3 Representation of Predicates in

FOL

This activity consists in translating the items

selected in the previous activity, expressed in natural

language to predicates in FOL. It consists of seven

sub-activities: “Identification of entities”,

“Redefinition of entities”, “Identification of

relationships”, “Redefinition of relationships”,

“Definition of the arity”, “Definition of predicates”

and “Redefinition of the entities of the predicate”

(Figure 2).

In the sub-activity “Identification of entities”, all

explicit or implicit subjects and objects in a sentence

are identified from the items selected in the previous

activity and exemplified in Table 1. The result of

this sub-activity is shown in Table 2.

Figure 2: “Representation of Predicates in FOL” activity.

Table 1: List of goals and facts of the ontology.

1 Calculate the inheritance of a person

2 Identify the heirs of a person

3 Determine the inheritance of each heir

4 Identify the spouse of a person

Define the inheritance portion of descendants in the

universal regime

6 Identify the ascendants of a person

7 Identify the descendants of a person

Determine the moiety of the spouse in the universal

regime

Determine the moiety of the spouse in the limite

regime

10 Determine the quantity of descendants of a person

11 Determine the quantity of ascendants of a person

Determine the portion of the descendants in the

universal regime

Determine the portion of the descendants in the

limited regime

14 Divide the moiety by the quantity of descendants

15 A person might have descendants

16 A person might have spouse

17 A person might have ascendants

18 A person might have marital properties

19 A person might have pre-marital properties

20 A person might have universal regime

21 A person might have partial regime

22 A marital property has value

23 A pre-marital property has value

Assign inheritance portion to spouse an

descendants in the universal regime

Assign inheritance portion to spouse an

descendants in the limited regime

26 Sum all the properties of a person

27 Verify the existence of ascendants

28 Verify the existence of descendants

29 Verify the existence of universal regime

30 Verify the existence of limited regime

The sub-activity “Redefinition of entities” takes

into account the entities identified in Table 2, and

for each one verifies if that is an entity or a

relationship. The entity “Heirs” is actually a

BuildingApplicationOntologiesthroughKnowledgeSystemGoals

117

relationship between two “People”, i.e., “A person is

heir of another person”. So, “Heirs” is redefined as a

“Person”, considering the entities that integrate the

relationship. However, the word “Heirs” is not

discarded, it will be useful in the sub-activity

“Redefinition of relationships”. Table 3 shows the

result of this sub-activity.

Table 2: Inputs and outputs of the sub activity

“Identification of entities’’.

Selected items Entities

Calculate the inheritance of a person

Inheritance,

Person

Identify the heirs of a person Heirs, Person

Determine the inheritance of each heir

Heirs,

Inheritance

Table 3: Inputs and outputs of the sub-activity

“Redefinition of entities”.

Entities Redefinition of entities

Inheritance, Person Inheritance, Person

Heirs, Person Person, Person

Heirs, Inheritance Person, Inheritance

The sub-activity “Identification of relationships”

uses the items selected in the activity “Selection of

Goals and Facts” to identify verbs in the phrases,

which represent the relationships to be extracted. For

instance, in the selected item “Calculate the

Inheritance of a person”, the verb “Calculate” is

identified as a relationship. Table 4 shows the

relationships identified.

Table 4: Inputs and outputs of the sub-activity

“Identification of relationships”.

Selected items Relationships

Calculate the inheritance of a person Calculate

Identify the heirs of a person Identify

Determine the inheritance of each heir Determine

The sub-activity “Redefinition of relationships”

takes into account the relationships identified in the

previous activity (exemplified in Table 4) and

verifies if these relationships are transitive verbs, as

they need a supplement to make sense. For example,

the relationship “identify” needs a supplement to

give it sense, using their respective entities identified

in Table 3 or the words that were considered entities

in the first sub-activity, for example, the word

“Heirs”. Table 5 shows the result of this sub-activity

applied to the examples in Table 3 and Table 4.

The sub-activity “Definition of the arity” defines

the number of entities involved in the relationships

previously identified. This quantity is determined

according to by the number of entities identified on

each selected item. Table 6 shows the arity identified

for the items 1, 2 and 3 in Table.

Table 5: Inputs and outputs of the sub-activity

“Redefinition of relationships”.

Relationships Entities

Redefinition of

relationships

Calculate Inheritance calculateInheritance

Identify Heirs identifyHeirs

Determine Inheritance determineInheritance

Table 6: Inputs and outputs of the sub-activity “Definition

of the arity”.

Relationships Entities Arity

calculateInheritance Inheritance, Person 2

identifyHeirs Person, Person 2

determineInheritance Person, Inheritance 2

In the sub-activity “Definition of predicates” the

entities and relationships identified and illustrated in

Tables 3 and 5 are represented in FOL. Table 7

presents the predicates resulting from the realization

of this sub-activity.

Table 7: An example of translation of the selected items

into predicates in FOL.

Selected items Predicates

Calculate the inheritance of a

person

calculateInheritance

(Person, Inheritance)

Identify the heirs of a person

identifyHeirs

(Person, Person)

Determine the inheritance of

each heir

determineInheritance

(Person, Inheritance)

The sub-activity “Redefinition of the entities of

the predicate” aims at renaming the arguments of the

predicates, defined in the sub-activity “Definition of

predicates”, when the arguments have the same

name. For example, for the predicate

“identifyHeirs(Person, Person)”, the entities are

considered variables since they represent distinct

persons. So it is redefined to “identifyHeirs

(PersonX, PersonY)” and this change is also

propagated to all other predicates in Table 7. Table 8

presents the result of this sub-activity and the final

product of this activity.

Table 8: Inputs and outputs of the sub-activity

“Redefinition of the entities of the predicate”.

Predicates Predicates redefined

calculateInheritance

(Person, Inheritance)

calculateInheritance

(PersonX, Inheritance)

identifyHeirs

(Person, Person)

identifyHeirs

(PersonX, PersonY)

determineInheritance

(Person, Inheritance)

determineInheritance

(PersonY, Inheritance)

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

118

2.4 Specification of Axioms in FOL

The purpose of this activity is to specify the rules

that lead to the achievement of the goals of the

system which are represented as predicates in FOL.

The process is iterative, because there is an iteration

with the activity “Selection of Goals and Facts”. For

each goal contained in a rule a search is performed

in the list of goals and facts to retrieve the subgoals

that satisfy it.

This activity consists of four sub-activities:

“Definition of the condition and conclusion”,

“Definition of Boolean operators”, “Definition of

quantifiers” and “Definition of implication or

equivalence”. Figure 3 shows the sub-activities of

this activity.

Figure 3: “Specification of axioms in FOL” activity.

The sub-activity “Definition of the condition and

conclusion” determines the condition and the

conclusion of each rule. The conclusion is the main

goal that has to be achieved and the condition can be

considered as a set of assumptions or subgoals that

lead to the achievement of the main goal. This sub-

activity receives as input the predicates identified in

Table 8. Table 9 shows the result of this sub-activity.

Table 9: Output of the sub-activity “Definition of the

condition and conclusion of the axiom”.

Condition and predicates of the

axiom

Conclusion

identifyHeir

(PersonX, PersonY)

calculateInheritance

(PersonX, Inheritance)

determineInheritance

(PersonY, Inheritance)

The sub-activity “Definition of Boolean

operators” specifies the Boolean operators which

integrate the predicates of the axiom condition. The

operators used are the conjunction represented by

the symbol (^) and the disjunction represented by the

symbol ().

Predicates in the condition are joined by an

“and” operator when all of them are needed to

achieve the conclusion; by an “or” operator when

they are alternative predicates to achieve the

conclusion. For example, to achieve the goal

“Calculate the inheritance of a person”

(calculateInheritance(PersonX, Inheritance)), it is

necessary to satisfy all the goals “Identify the heirs

of a person” (identifyHeirs(PersonX, PersonY)) and

“Determine the inheritance of each heir”

(determineInheritance(PersonY,Inheritance)).

Therefore, a conjunction is used to integrate these

two predicates.

The sub-activity “Definition of quantifiers”

defines the appropriate quantifiers associated to

entities present in the axiom. Quantifiers can be

universal () or existential (). The first one is used

to indicate that a predicate is true for all the elements

of a given set while the last one is used to indicate

that a predicate is true for at least one element in a

given set. For instance, the variable “PersonX”

refers to “at least one person who died” so the

existential quantifier is associated to this entity. The

variables “PersonY” and “Inheritance” follow the

same principle, being set to the existential quantifier.

The sub-activity “Definition of implication or

equivalence” takes as input a set of predicates like

those in the example of Table 9 and to determines

whether the axiom to be created is an implication or

an equivalence.

The implication is used when the satisfaction of

the condition leads to the conclusion. The

equivalence occurs when there is a symmetry

between the condition and conclusion. For instance,

an implication is used to form the following axiom:

“PersonX, PersonY, Inheritance | identifyHeirs

(PersonX,PersonY) ^ determineInheritance

(PersonY,Inheritance) calculateInheritance

(PersonX,Inheritance)”.

As illustrated in Figure 1, the GAODT activities

are executed iteratively. Therefore, in order to

construct new axioms, each one of the predicates in

the condition of the current axiom submitted to the

“Selection of Goals and Facts” activity where once

more time the items in Table, satisfying this

condition will be selected.

For instance, the predicate goal “identifyHeirs

(PersonX,PersonY)” which is part of the condition

in the axiom example of the previous paragraph is

submitted to the “Selection of Goals and Facts” and

the domain specialist informs that the items “Verify

the existence of descendants”, “Identify the

descendants of a person”, “Verify the existence of

spouse”, “Identify the spouse of a person” satisfy

that goal. Then, new goals specified in natural

language are given as input to the activity

“Representation of Predicates in FOL” (Table 10)

BuildingApplicationOntologiesthroughKnowledgeSystemGoals

119

and a new cycle of the GAODT technique begins.

Table 10: Goals represented as predicates.

Goals in Natural Language Predicates

Verify the existence of

descendants

verifyExistence

Descendants

(PersonX, PersonY)

Identify the descendants of a

person

identifyDescendants

(PersonX, PersonY)

Verify the existence of spouse

verifyExistenceSpouse

(PersonX, PersonY)

Identify the spouse of a person

identifySpouse

(PersonX, PersonY)

These predicates are given as input to the activity

“Specification of Axioms in FOL” to generate the

new axiom presented in Table 11.

Table 11: Axiom developed in the activity “Specification

of Axioms in FOL”.

CONDITION

verifyExistenceDescendants(PersonX, PersonY) ^

identifyDescendants(PersonX, PersonY) ^

verifyExistenceSpouse(PersonX, PersonY) ^

identifySpouse(PersonX, PersonY)

CONCLUSION

identifyHeirs(PersonX, PersonY)

The predicate “determineInheritance(PersonY,

Inheritance)” that also makes part of the axiom

condition pass through the same process of

specification and representation in FOL to which the

predicate “identifyHeirs(PersonX,PersonY)” was

submitted, generating the new axiom in Table 12.

Table 12: Axiom developed in activity “Specification of

Axioms in FOL”.

CONDITION

verifyExistenceDescendants(PersonX, PersonY) ^

verifyExistenceSpouse(PersonX, PersonY) ^

verifyExistenceUniversalRegime

(PersonX,UniversalRegime) ^

assignInheritancePortionDescendants

SpouseUniversalRegime

(PersonX Inheritance,

PersonY, UniversalRegime)

CONCLUSION

determineInheritance(PersonX,PersonY)

The process is then recursively executed for each

one of the subgoals, until all the goals given as input

to the technique (as the ones illustrated in Table 1)

have been satisfied. The product of this activity is a

set of axioms specified in predicates in FOL. A sub-

set of the axioms generated from the activity

“Specification of Axioms in FOL” is shown in Table

13.

2.5 Specification/Extension of the

Application Ontology

The constituent elements of the axioms specified in

the previous activity are extracted for the

construction of the application ontology. This

activity consists of six subactivities: “Translation of

axioms”, “Definition of classes”, “Definition of non-

taxonomic relationships”, “Definition of taxonomic

relationships”, “Definition of properties” and

“Retrieval of application ontologies” (Figure 4).

Table 13: A sub-set of axioms generated from the activity “Specification of Axioms in FOL.

Condition Conclusion

identifyHeirs(PersonX, PersonY) ^

determineInheritance(PersonX, Inheritance)

calculateInheritance

(PersonX, Inheritance)

verifyExistenceDescendant(PersonX, PersonY) ^

verifyExistenceSpouse(PersonX, PersonY) ^

verifyExistenceUniversalRegime(PersonX, UniversalRegime) ^

assignInheritancePortionDescendantsSpouseUniversalRegime

(PersonX, Inheritance, PersonY, UniversalRegime)

determineInheritance

(PersonX, PersonY)

mightHaveDescendants(PersonX, PersonY)

verifyExistenceDescendants

(PersonX, PersonY)

mightHaveSpouse(PersonX, PersonY)

verifyExistenceSpouse

(PersonX, PersonY)

verifyExistenceDescendants(PersonX, PersonY) ^

identifyDescendants(PersonX, PersonY) ^

verifyExistenceSpouse(PersonX, PersonY) ^

identifySpouse(PersonX, PersonY) 

verifyExistanceAscendants(PersonX, PersonY) ^

identifyAscendants(PersonX, PersonY)

identifyHeirs

(PersonX, PersonY)

determinePortionDescendantsUniversalRegime(PersonX, Portion) ^

determineMoietySpouseUniversalRegime(PersonX, PersonY)

assignInheritancePortionDescendants

SpouseUniversalRegime

(PersonX, Inheritance,

PersonY, UniversalRegime)

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

120

Figure 4: “Specification/extension of the application

ontology” activity.

The sub-activity “Translation of axioms”

converts the axioms defined in the previous activity

expressed in FOL into rules expressed in an

ontology rule based language, like RuleML (Harold,

2001). The experiences conducted to evaluate

GAODT use RuleML because of its expressiveness.

To perform this translation, the following

heuristics are applied. First, regular expressions

(Jeffrey, 2001) are used to extract the premises and

the conclusions of the axioms. For example, for the

rule “PersonX, PersonY, Inheritance |

identifyHeirs(PersonX,PersonY) ^ determine

Inheritance(PersonY,Inheritance) calculate

Inheritance(PersonX,Inheritance)”, the following

regular expression was used “^(\w+$.*$) ^

(\w+$.*$)  (\w+$.*$)”. Then, the premises and

conclusion are specified in POSL (Boley, 2004) and

finally automatically translated to RuleML axioms

(Figure 5).

The sub-activity “Definition of classes” extracts

the variables of the axioms of the “Specification of

Axioms in FOL” activity illustrated in Table 13. For

example, the predicate “identifyHeirs

(PersonX,PersonY)” has the variables “PersonX”

and “PersonY” both referring to the class “Person”.

The sub-activity “Definition of non-taxonomic

relationships” extracts non-taxonomic relationships

of the ontology from the predicates in the list of

axioms outputted from the activity “Specification of

Axioms in FOL” (Table 13). A non-taxonomic

relationship is defined for each predicate having the

same name and arity. For example, the predicates

“sumProperties(PersonX,Properties)”, “sumProperti-

es(PersonX,MaritalProperty)” and “sumProperties

(PersonX,Pre-maritalProperty)” have the same name

and arity; therefore, the non-taxonomic relationship

“sumProperties/2” is defined.

<And>

<Atom>

<Rel> identifyHeirs </Rel>

<Var>PersonX</Var>

<Var>PersonY</Var>

</Atom>

<Atom>

<Rel>determineInheritance</Rel>

<Var>PersonX</Var>

<Var>PersonY</Var>

</Atom>

</And>

<Atom>

<Rel>calculateInheritance</Rel>

<Var>PersonX</Var>

<Var>PersonY</Var>

</Atom>

</Implies>

</Rulebase>

</Assert>

Figure 5: Example of an axiom represented in RuleML.

The sub-activity “Definition of taxonomic

relationships” extracts a set of taxonomic

relationships based on the hierarquical relation

between the variables of the predicates outputted

from the previous sub-activity. For example, there is

a hierarchy between the classes, “Property”,

“MaritalProperty” and “Pre-maritalProperty”,

extracted from the predicates “sumProperties

(PersonX,Properties)”, “sumProperties(PersonX,

MaritalProperty)” and “sumProperties(PersonX,Pre-

maritalProperty)”.

The sub-activity “Definition of properties”

extracts from the axioms the predicates describing

attributes of the classes. For example,

“hasValue(MaritalProperty,Value)” and “hasValue

(Pre-maritalProperty,Value)” describe that the

classes “MaritalProperty” and “Pre-maritalProperty”

has the property “hasValue”.

Figure 6: Taxonomic relationships of the Intestate

Succession application ontology.

BuildingApplicationOntologiesthroughKnowledgeSystemGoals

121

Table 14: A sub-set of the non-taxonomic relationships of

the Intestate Succession application ontology.

Domain Relationship Range

Person calculateInheritance Inheritance

Person identifyAscendants Person

Person identifySpouse Person

Person identifyDescendants Person

Person determineInheritance Inheritance

Person identifyHeirs Person

Person mightHaveDescendants Person

Person mightHaveMarital MaritalProperty

Person

mightHave

Pre-Marital

Property

Person

mightHave

UniversalRegime

Universal Regime

Person

mightHave

LimitedRegime

Limited

Regime

Table 15: A sub-set of the properties of the Intestate

Succession application ontology.

Domain Properties

Person determineQuantityAscendants

Person determineQuantityDescendants

Person sumMaritalProperties

Person

determineMoietySpouse

UniversalRegime

Person

determinePortionDescendants

UniversalRegime

Person

determinePortionDescendants

LimitedRegime

Person

determineMoietySpouse

LimitedRegime

Pre-

MaritalProperty

hasValue

MaritalProperty hasValue

<And>

<Atom>

<Rel>hasValue</Rel>

<Var>MaritalProperty</Var>

<Var>Value</Var>

</Atom>

<Atom>

<Rel> sumMaritalProperties</Rel>

<Var>Person</Var>

<Var> MaritalProperty </Var>

</Atom>

</And>

</Implies>

</Rulebase>

</Assert>

Figure 7: Example of an axiom represented in RuleML.

If there is a need to extend the application

ontology developed, the sub-activity “Retrieval of

application ontologies” performs a semantic search

for reusable ontologies in a repository. Several

similarity measures (Lee et al., 2008); (Claudia et

al., 2008) can be used to rank the ontologies

retrieved.

Figure 6, Figure 7, Table 14 and Table 15 show

respectively the taxonomy, an example of an axiom,

a sub-set of the non-taxonomic relationships and

properties of the application ontology developed for

the domain of Intestate Succession, which make up

the final product of GAODT.

3 RELATED WORK

Various techniques and methodologies for building

ontologies have been proposed like the 101

technique (Noy and McGuinness, 2001),

DERONTO (Caliari, 2007), Uschold and King

(Uschold and King, 1995) (Fernandez et al., 2004),

Gruninger and Fox (Gruninger and Fox, 1995);

(Fernandez et al., 2004) and Methontology (Perez,

2004); (Fernandez et al., 2004). Table 16 shows the

results of a comparative analysis among them and

GAODT technique according to the following

criteria: the type of developed ontology, the order

through which the ontology elements are discovered,

the type of life cycle (classic or iterative), the reuse

of existing ontologies and the use linguistic

knowledge to find classes and relationships.

Ontologies can be of four types: high-level,

domain, task and application ontologies (Guarino,

1998). The chosen technique should be appropriate

for the construction of a particular type of ontology.

Among the state of the art of the techniques

analyzed in this work only GAODT and DERONTO

support the development of application ontologies.

An application ontology is composed of six

elements: classes, taxonomy, relationships,

properties, axioms and instances (Girardi, 2010).

The order in which these elements are discovered

influences the development of the subactivities of

the technique or methodology. For example,

GAODT starts by discovering axioms considering

that they represent the requirements of the

knowledge-based system. Other techniques begin

with the identification of relevant domain terms

corresponding to ontology classes.

An ontology development process should

preferably be incremental allowing the engineer to

add new elements to the ontology at each process

interaction. DERONTO, 101, Methontology and

GAODT are supported by an incremental

development process.

Good techniques and methodologies should

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

122

Table 16: Comparative analysis of techniques for the construction of ontologies.

Comparison criteria DERONTO

Uschold and

King

101 Methontology

Gruninger and

Fox

GAODT

The type of developed

ontology

Domain and

application

ontologies

Domain

ontologies

Domain ontologies

Domain

ontologies

Domain

ontologies

Application

ontologies

The order through

which the ontology

elements are

discovered

Classes

Taxonomy

Properties

Relationships

Axioms

Classes

Taxonomy

Relationships

Classes

Taxonomy

Properties

Relationships

Axioms

Instances

Classes

Taxonomy

Relationships

Properties

Axioms

Instances

Classes

Relationships

Properties

Instances

Axioms

Taxonomy

Axioms

Classes

Relationships

Taxonomy

Properties

The type of life cycle Iterative Classic Iterative Iterative Classic Iterative

The reuse of existing

ontologies

No Yes Yes Yes No Yes

Uses linguistic

knowledge

No No Yes No No Yes

consider the reuse of existing ontologies like domain

and task ontologies and even application ontologies.

Most of the analyzed techniques approach

reusability with the exception of DERONTO and

Gruninger and Fox.

Only GAODT and the 101 technique define rules

to find classes from nouns and relationships from

verbs. This is an advantage because it allows easy

identification of the elements of the ontology based

on a linguistic strategy. Another advantage of the

use of linguistic knowledge is the possible

automation of the extraction of these elements using

NLP techniques (Cunningham et al., 2012).

4 CONCLUSIONS AND FUTURE

WORK

This article described GAODT, a technique for

building application ontologies through a goal-

oriented development cycle. The technique also

provides the developer, a well-defined way to

translate the knowledge expressed in natural

language to FOL predicates. This feature is not

covered by any other technique or methodology of

the state of the art presented in this paper.

GAODT has been evaluated through the

development of a study case for the construction of

an application ontology to be used on an Intestate

Succession Knowledge-based System.

Building reusable ontologies is a costly process.

Among the four types of ontologies defined by

Guarino (Guarino, 1998), application ontologies are

the less reusable once they are developed for

specific software applications. However, they are

generally easier and faster to develop. Building

application ontologies and then generalizing its

elements to domain and task ontologies is a good

alternative approach for developing these reusable

artifacts. In this context, GAODT consists of a first

step in this direction by defining a systematized way

for building application ontologies.

Future improvements of GAODT include the

development of a software tool to support it and a

technique to perform the semantic search for

ontologies to be reused. GAODT will also be

integrated into a knowledge based process for the

development of multi-agent systems (Leite, 2009)

already constructed by the authors´ research group.

The main objective is to use GAODT to construct

the knowledge bases of deliberative agents of

knowledge-based systems developed with this

process.

ACKNOWLEDGEMENTS

This work is supported by CNPq, CAPES and

FAPEMA, institutions of Brazil and Maranhão

Governments for scientific and technologic

development.

REFERENCES

Article nº 1.829., 2002. Brazilian Civil Code. Federal

Legislation.

Boley, H., 2004. POSL: An Integrated Positional-Slotted

Language for Semantic Web Knowledge. W3C.

Caliari, F., 2007. DERONTO: Method for Building

Ontologies from Entity-Relationship Diagrams.

Master thesis - Federal Technological University of

Paraná. (In Portuguese)

Claudia, A., Fanizzi, N., Esposito, F., 2005. A semantic

similarity measure for expressive description logics. In

BuildingApplicationOntologiesthroughKnowledgeSystemGoals

123

Proceedings of Italian Conference on Computational

Logic, Roma.

Costa, A., 2009. MADAE-Pro A knowledge-based process

for Domain and Application Engineering. Master

thesis - Federal University of Maranhão. In

Portuguese.

Cunningham, H., Maynard, D., Bontcheva, K., Tablan, V.,

2002. GATE: A Framework and Graphical

Development Environment for Robust NLP Tools and

Applications, In: Proceedings of the 40th Anniversary

Meeting of the Association for Computational

Linguistics (ACL'02). Philadelphia, July.

Fernández, M., Pérez, A., Juristo, N., 1997. Methontology:

From Ontological Art Towards Ontological

Engineering. Spring Symposium Series. Stanford.

Girardi, R., 2010. Guiding Ontology Learning and

Population by Knowledge System Goals. In:

Proceedings of the International Conference on

Knowledge Engineering and Ontology Development,

Ed. INSTIIC, Valence, p. 480 – 484.

Gonçalves, C., 2009. Brazilian Civil Law: Inheritance

law. São Paulo, Saraiva. (In Portuguese).

Gruninger, M., Fox, M., 1995. Methodology for the design

and evaluation of ontologies. In: IJCAI95 Workshop

on Basic Ontological Issues in Knowledge Sharing,

Montreal, Canada.

Guarino, N., 1998. Formal Ontology in Information

Systems. Proceedings of the 1st International

Conference, Trento, Italy, IOS Press.

Harold, B., 2001. The Rule Markup Language: RDF-XML

Data Model, XML Schema Hierarchy, and XSL

Transformations. In Proc. 14th International

Conference on Applications of Prolog.

Jeffrey, F., 2006. Mastering Regular Expressions. O'Reilly

Media, 3rd Edition.

Lee, W., Shah, N., Sundlass, K.., Musen, M., 2008.

Comparison of ontology-based semantic-similarity

measures. AMIA Symposium, p. 384–388.

Noy, N., McGuinness, D., 2001. Ontology Development

101: A Guide to Creating Your First Ontology.

Pérez, A., et al., 2004. Ontological Engineering: with

examples from the areas of knowledge management, e-

commerce and the semantic web. London. Springer-

Verlag.

Russel, S., Norvig, P., 2004. Artificial Intelligence. Rio de

Janeiro: Ed. Campus. (In Portuguese).

Uschold, M., King, M., 1995. Towards a Methodology for

Building Ontologies. In IJCAI’95 Workshop on Basic

Ontological Issues in Knowledge Sharing. Montreal,

Canada.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

124