Resumen

La amplia aplicabilidad del mapeo de reglas de negocio en sentencias ontológicas ha sido recientemente reconocida. Algunas de las aplicaciones más importantes son: (1) la utilización de razonadores ontológicos para probar la consistencia de la información del dominio del negocio, (2) la generación de una ontología destinada a ser utilizada en la etapa de análisis del proceso de desarrollo de software, y (3) la posibilidad de encapsular la especificación declarativa del conocimiento del negocio en un sistema de información mediante la implementación de una ontología. El lenguaje denominado Semantics of Business Vocabulary and Business Rules (SBVR) brinda soporte a tal enfoque al proveer al experto del negocio una forma lingüistica de describir semánticamente los conceptos y especificar las reglas de negocio, en forma independiente al diseño del sistema de información. Aunque trabajos previos han presentado algunas propuestas, resta definir un enfoque exhaustivo y automatizable de tales mapeos. Este trabajo presenta un conjunto amplio y detallado de transformaciones que permiten la generación automatizable de una ontología implementada en OWL 2, partiendo de la especificación SBVR del dominio del negocio. Tales transformaciones están basadas en la especificación estructural de ambos estándars, y son descriptas mediante un caso de estudio. Se ha realizado un ejemplo real de validación, abordando la factibilidad de los mapeos mediante la evalución de la calidad de la ontología resultante.

Palabras clave: regla de negocio, ontología, mapeo, SBVR, OWL 2

Received: 2013-09-09, Revised: 2014-02-05, Accepted: 2014-02-05

1 Introduction

The Semantics of Business Vocabulary and Business Rules (SBVR) provides business people a linguistic way to semantically describe business concepts and specify business rules [1]. The linguistic approach adopted by the proposal enables the expression of business knowledge through statements rather than diagrams. That is based in the insight that diagrams are helpful for depicting structural organization of concepts but they are impractical as a primary means of defining vocabularies and expressing business rules.

SBVR is rooted in first-order predicate logic with some restricted extensions into higher-order logics and some limited extensions into modal logic. Despite such sound theoretical foundation on formal logic is a key feature in automated reasoning contexts, SBVR has been conceptualized for business people and designed to be used for business purposes independent of information system designs.

Several ]]> works have already recognized the benefits of having a mean of mapping SBVR expressions to ]]> on- tology statements. Important applications of such mappings can be mentioned. For example, ontology reasoners ]]> could be used to automatically prove the consistency of business models [2][3][4][5]. Ontologies intended to be ]]> used in the analysis stage of a software development process could be generated ]]> from main business knowledge sources [6]. The mappings could also be used to generate ontologies that ]]> encapsulate business knowledge into information software systems, enabling unambiguous representation of knowledge and efficient management ]]> of highly dynamic environments [7][8][9]. Use of semantic technologies for ]]> creating more intelligent and effective enterprise information systems has increased considerably in the last years. Several works have ]]> already recognized that having a mean of mapping SBVR expressions to ontology state- ments ]]> will have strong benefits and will gain wide applicability in ]]> the future vision of enterprise information systems [2][7][3][4][5].

Although previous proposals have presented some SBVR to OWL 2 transformations [2][7][3][4][5], an ex- haustive and automatable approach for them is still lacking. Some mappings between SVBR and OWL 1 are presented in [2] through an example, but the transformations are not explicitly formalized and - as a consequence - they can not be generalized to other situations. The proposal depicted in [7] presents a set of transformations based on OWL 1, rooted in the insight that SBVR rules have to be mapped into a rule language while OWL is used just for describing the vocabulary which the rules build on. Another set of transformations is presented in [3], but just some primary questions are answered. Meanwhile, [4] and [5] present a set of transformations rooted in the ORM conceptual modelling language, but the proposal follows a formally grounded approach by stating the logical foundations of the translations between the underlying theories of SBVR and OWL 2.

This work presents a broad and detailed set of transformations that allows the automatable generation of an OWL 2 ontology from the SBVR specifications of a business domain. Transformations are rooted on the structural specification of both standards and are depicted along the paper through a case study. OWL 2

Web Ontology Language (OWL 2) has been selected as the receipt language of the transformations because it has evolved as a de-facto standard for a broad spectrum of applications [10].

The rest of this paper is organized as follow. Section 2 analyses the differences with previous works. Sec- tion 3 and Section 4 provides an overview of the SBVR and OWL 2 specifications, respectively. Section 5 presents the SBVR to OWL 2 mappings and illustrate them through a case study. Section 6 depicts a real case validation example performed by a group of undergraduate engineering students. Section 7 presents some discussions about the addressed topics. Finally, some conclusions and future research directions are presented in Section 8.

2 Related Work

In literature, it can be found that several works have already recognized that having a mean of mapping SBVR expressions to ontology statements will have strong benefits and will gain wide applicability in the future vision of enterprise information systems [2][7][3][4][5].

An approach combining OWL 1 [11] and SWRL expressions [12] is explored in [2]. The aim of the approach is to obtain a Platform Independent Model (PIM) [13] from the SBVR business vocabulary and rules ex- pressions. That goal is achieved by performing an intermediate SBVR to OWL/SWRL translation process as a way to check consistency and expand the knowledge by means of inference procedures. The work ex- plores the feasibility of the proposed solution and the main problems arising. Although it illustrates some mappings between SVBR and OWL 1 through an example, the transformations are not explicitly formalized and therefore they can not be generalized to other situations.

A model transformation chain is presented in [7] as a way to translate SBVR based vocabularies to OWL

1 and R2ML statements (REWERSE Reasoning ]]> on the web: R2ML The REWERSE I1 Rule Markup Language. Accessible in: http://goo.gl/p6kpjR) . Conducted by the insight that ]]> business rules should be guaranteed by all IT applications of an enterprise, the paper presents the needs ]]> for a technology to transform the SBVR expres- sions into a PIM. ]]> The work meets such need by presenting a chain of metamodel-based ]]> transformations from a conceptual point of view. The transformations are rooted on an intermediate metamodel of SBVR and the QVT language [13]. ]]> The proposal use SBVR Vocabulary-to-MOF/XMI Mapping Rule Set to produce a MOF model and XML schema [13], ]]> the ODM [14] in order to describe OWL models within the MOF context, and the MOF-based metamodel of R2ML to generate the final version of the rules. The authors highlight ]]> the contribution to the model-driven integrity engineering of their approach given that the model transformation ]]> chain is completely based on the abstract syntax of the involved languages.

Some basics concepts and problems in transforming SBVR expressions to OWL 2 are introduced in [3]. The aim of the work is to allow business users to describe ontologies in a similar way to their everyday business language, which will allow to prove consistency of business vocabulary and rules by means of OWL 2 reasoners. Although this proposal also show the feasibility of using SBVR expressions for the conceptual modeling of an ontology, just answer some primary questions.

On the other hand, [4] and [5] propose a set of transformations but following a different approach. These works are rooted in the ORM conceptual modelling language (Object Role Modeling. The official site for conceptual data modeling. Accessible in: http://orm.net/), which is at the core of the SBVR proposal. Definition and application of an integrated method that uses ORM to generate an ontology-based query mechanism is presented in [4]. Finally, a set-theoretic semantic for ORM 2 and a formal approach to map a fragment of such language to the ALCQI fragment of Description Logics [16] is introduced in [5]. Although the work also depicts a tool which implements the translations of a set of ORM 2 constraints into a OWL 2 ontology, the proposal follows a formally grounded approach by stating the logical foundations of the translations between the underlying theories of SBVR and OWL 2.

Different from previous work, the present paper depicts a broad and detailed set of transformations that allows the automatable generation of an OWL 2 ontology from the SBVR specifications of a business do- main. Transformations are rooted on the structural specification of both standards rather than theoretic considerations of the language, with the aim to provide a set of mappings readily usable for business people or developers concerned with the implementation of a mapping tool.

3 SBVR Overview

SBVR defines the vocabulary and rules for documenting the semantics of business vocabularies, business facts, and business rules; which allows their verbalization in a controlled vocabulary readily understandable by business people.

The fact-oriented approach of SBVR stems from the Business Rules Manifesto (http://www.businessrulesgroup.org/brmanifesto.htm), stating that rules builds on facts, and facts build on concepts as expressed by terms. Therefore, terms express business concepts, facts make assertions about these concepts, and rules constrain and support these facts. SBVR supports such approach by providing noun concepts and verb concepts respectively corresponding to the notions of terms and facts. Figure 1 shows the structural organization of such components.

A verb concept - also named fact type - is a concept that is the meaning of a verb phrase that involves one or more noun concepts. It can be used to represent unary relations - i.e., unary fact type -, binary relations - i.e., binary fact type and even n-ary relations. Another important kind of fact type is the Specialization, which represents a relationship between a more general and a more specific concept. Moreover, a Reference Scheme is the SBVR way of identifying instances of a given concept by means of its characteristics specified by means of one or more unary fact types.

Figure 1: Noun and verb concepts in SBVR

Finally, a SBVR rule is an element of guidance that introduces an obligation or necessity and distinguishing two general types: (1) structural rules, which describe the way the business chooses to organize the things it deals with; and (2) operative rules, which govern the conduct of business activity by describing business processes. Both types of rules are built imposing restrictions over fact types by using quantifiers, logical operators, etc. Figure 2 shows the structural organization of quantifications and logical operations.

As early stated, SBVR adopts a linguistic approach that allows to define vocabularies and express operative rules. According to this insight, SBVR defines a Controlled Natural Language (CNL) - named SBVR Structured English - and describes the way to mechanically mapping such CNL expressions to SBVR formal concepts.

4 OWL 2 Overview

The OWL 2 Web Ontology Language (OWL 2) is the latest version of an ontology language proposed by the World Wide Web Consortium (W3C) for the development of the Semantic Web [17], but it has gradually evolved as a de-facto standard for a broad spectrum of applications.

Figure 2: Quantifications and logical operations in SBVR

OWL 2 ontologies provide classes, properties, individuals, and data values, and are stored as Semantic Web documents. An OWL 2 ontology is a formal description of a domain of interest rooted in three syntactic categories that are interpreted under a standardized semantics, which allows useful inferences to be drawn. Figure 3 , Figure 4 and Figure 5 depict the structural specification of such categories:

• Entities such as classes, properties, and individuals. They are the basic elements of an ontology and are identified by Internationalized Resource Identifiers (IRIs) [18]. For example, a class a:Person can be used to represent the set of all people, the object property a:parentOf can be used to represent the parent-child relationship and the individual a:Peter can be used to represent a particular person called

“Peter”.

Figure 3: Entities in OWL 2 ontologies

• Expressions, representing complex notions in the domain being described. For example, a class expres- sion describes a set of individuals in terms of the restrictions on the individuals characteristics.

• Axioms, which are statements asserted to be true in the domain being described. For example, a subclass axiom state that the class a:Student is a subclass of the class a:Person.

OWL 2 ontology language defines several concrete syntaxes that can be used to serialize and exchange ontologies. Among them, the functional style syntax is defined in the OWL 2 structural specification [17] with the aim to state the semantics of OWL 2 constructors and allow a compact writing of ontologies. Following the structural specification insight, the rest of this paper uses such syntax to state OWL 2 ontology expressions.

Figure 4: Expressions in OWL 2 ontologies

Figure 5: Axioms in OWL 2 ontologies

It is important to highlight that the latest version of OWL present new features although it has a very similar overall structure to OWL 1 and backwards compatibility is complete - i.e., all OWL 1 ontologies remain valid OWL 2 ontologies with identical inferences in all practical cases -. While some of such new features can be seen as syntactic sugar, there is a wide set of added constructors providing more expressive capabilities such as keys, property chains, richer datatypes, data ranges, qualified cardinality restrictions, and asymmetric, reflexive, and disjoint properties. A document detailing the new features can be found in [15].

Mappings presented in this section allow the automatable generation of an OWL 2 ontology from the SBVR specifications of a business domain. Transformations are rooted on the structural specification of both standards and are depicted in subsections below by grouping and sequencing them according to the inherent logical order of the subject matter itself. In addition to their theoretical expression, the mappings are illustrated by building an ontology that reflects the business knowledge exposed by a case study (Ontology generated from the example can be found in http://code.google.com/p/eurent-mapping-case-study/).

The case study has been drawn from the Annex E of the SBVR specification (Annex E (Pp. 267) of the SBVR specification. Accessible in: http://www.omg.org/spec/SBVR/1.0/), depicting the business service

of a fictitious car rental company with ]]> branches in several countries. This case study has been introduced by the OMG as a ]]> common real world model of business enterprise that researchers and system developers ]]> can use to illustrate the capabilities of their proposals. In this case study, ]]> the ontological conceptualization of the business knowledge provide a way to ]]> automatically prove the consistency of the business model [2][3][4][5]. Although just a ]]> fragment of the case study is presented, it is complete ]]> enough in order to depict the usefulness of the proposed mappings. Figure 6 exposes the movements of cars ]]> among branches used to illustrate the mappings. EU-Rent rents cars to its customers. Different models ]]> of car are offered, organized into groups. All cars in a group ]]> are charged at the same rates. A rental booking specifies the car ]]> group required, the start and end dates/times of the rental and the EU-Rent ]]> branch from which the rental is to start. Optionally, the reservation may specify a one ]]> way rental - in which the car is returned to a branch different ]]> from the pick up branch - and may request a specific car model within the required group. Furthermore, the rentals are classified ]]> based on whether it crosses local area or international boundaries.

Figure 6: EU-Rent car movement among branches

5.1 Core Mappings

As early stated, object types, individual concepts, fact types and fact type roles constitute the core of SBVR

metamodel. Therefore, their mappings to OWL 2 expressions are required for more complex translations:

1. Each object type ot is mapped to Declaration(Class(a:ot))

2. Each individual concept ic of an object type ot is mapped to:

Declaration(NamedIndividual(a:ic) ClassAssertion(a:ot a:ic)

3. Each unary fact type uft is mapped to: Declaration(DataProperty(a:uft) DataPropertyDomain(a:uft a:ClassOne) DataPropertyRange(a:uft a:DataRangeOne)

4. Each binary fact type bft is mapped to: Declaration(ObjectProperty(a:bft) ObjectPropertyDomain(a:bft a:ClassOne) ObjectPropertyRange(a:bft a:ClassTwo)

5. Each fact type role ftr is mapped by using the SubObjectPropertyOf and ObjectPropertyChain OWL 2 axioms, as proposed in [3].

Table 1: SBVR specification of “car movement” concept

Table 2: SBVR specification of “branch” concept

Table 3: Mapping of main object types

3. at-most-n quantification, where “n” is a non-negative integer

If the logical formulation scopes over a unary fact type, the expression is mapped to:

DataMaxCardinality(n a:DataPropertyOne a:DataRangeOne)

ObjectMaxCardinality(n a:ObjectPropertyOne a:ClassOne)

4. at-least-n quantification, where “n” is a non-negative integer

If the logical formulation scopes over a unary fact type, the expression is mapped to:

DataMinCardinality(n a:DataPropertyOne a:DataRangeOne)

If the logical formulation scopes over a binary fact type, the expression is mapped to:

ObjectMinCardinality(n a:ObjectPropertyOne a:ClassOne)

5. exactly-n Quantification, where “n” is a non-negative integer

DataExactCardinality(n a:DataPropertyOne a:DataRangeOne)

If the logical formulation scopes over a binary fact type, the expression is mapped to:

ObjectExactCardinality(n a:ObjectPropertyOne a:ClassOne)

Remaining SBVR quantifications - at-most-one, exactly-one, and numeric-range quantifications - are easily translatable in terms of the above presented mappings.

Two exactly-n quantification mapping (where n = 1) are shown in Table 12 , Table 13 , Table 14 and Table 15 . However, the last one is stated as a numeric-range quantification with the aim of demonstrate such mapping. In this way, the exactly-1 quantification is expressed as a numeric-range quantification with upper and lower limit fixed at the value 1.

Table 12: SBVR specification of “car movement has receiving branch” fact type

Table 13: SBVR specification of “car movement has sending branch” fact type

Table 14: Mapping of exact quantification

Table 15: Mapping of exact quantification

5.3 Logical Operations Mappings

SBVR defines logical operations as those formulations of meaning based on only the truth or falseness of the meanings of its logical operands. In correspondence with the previous translations, the mappings of the SBVR logical operations are presented depending on the types of the involved logical operands:

1. Logical Negation

If the logical operand is an object type, then the expression is mapped to:

ObjectComplementOf(a:operand)

If the logical operand is a literal, then the expression is mapped to:

DataComplementOf(a:operand)

2. Conjunction

If both logical operands are object types, then the expression is mapped to:

ObjectIntersectionOf(a:operand1 a:operand2)

If both logical operands are literals, then the expression is mapped to:

DataIntersectionOf(a:operand1 a:operand2)

3. Disjunction

If both logical operands are object types, then the expression is mapped to:

ObjectUnionOf(a:operand1 a:operand2)

DataUnionOf(a:operand1 a:operand2)

4. Equivalence

If both logical operands are object types, then the expression is mapped to:

EquivalentClasses(a:operand1 a:operand2)

If both logical operands are individual concepts, then the expression is mapped to:

SameIndividual(a:operand1 a:operand2)

If both logical operands are unary fact types, then the expression is mapped to:

EquivalentDataProperties(a:logicaloperand1 a:logicaloperand2)

EquivalentObjectProperties(a:operand1 a:operand2)

Remaining SBVR logical operations - exclusive disjunction, nand-formulation, nor formulation, and whether- or-not formulation - are translatable by the logical combination of the above presented mappings.

As an example, the mapping shown in Table 16 depicts a disjunction expression stating that a rental in which the car is returned to a branch different from the pick up branch can crosses local area or international boundaries.

Table 16: Mapping of disjunction.

5.4 Identifiers, Specializations and Classification Mappings

1. Reference Scheme, is the SBVR way of identifying instances of a given concept. However, SBVR con- siders only characteristics to be used as reference schemes, so the expression is mapped to: HasKey(a:ClassExpression a:DataPropertyExpressionOne)

In the case study, the has Movement-ID characteristic is used as the identifier of the individuals be- longing to the Car Movement object type. The corresponding translation is:

HasKey(eurent:Car Movement () (eurent:has MovementID))

2. Specialization, is a fact type representing relationships between a more general and a more specific concept.

If both concepts are object types, then the expression is mapped to:

SubClassOf(a:concept1 a:concept2)

If both concepts are unary fact types, then the expression is mapped to:

SubDataPropertyOf(a:concept1 a:concept2)

SubObjectPropertyOf(a:concept1 a:concept2)

If concept1 is an object type and concept2 is an individual concept, the expression is mapped to:

ClassAssertion(a:concept1 a:concept2)

An application of the specialization mapping over the case study has been shown in the disjunction expression stated in the previous subsection.

3. Categorization and Segmentation

A possible situation is the need of categorizing the individuals belonging to certain entity according to a set of different criteria. Such situation is presented in the case study, where the car movements are classified based on whether the car is returned to a branch different from the pick up branch - directional categorization criteria - and on whether the car crosses local area or international bound- aries - geographical categorization criteria -. Moreover, the categories belonging to each criteria are disjoint between them. In the SBVR context, such modelling are called categorization and segmenta- tion, respectively. While OWL 2 ObjectUnionOf is the way to map a SBVR categorization, OWL 2

DisjointUnion are used to translate SBVR segmentations.

Multiple categorization criteria have been presented in the case study description through the rentals classifications based on directional or geographical aspects. The translations of such business issue are presented in Table 17 .

Table 17: Mapping of disjunction

6 A Real Case Validation Example