Many variants of the i* Framework have been developed since its definition. For instance, i* -based modeling languages have been proposed for agent-oriented and service-oriented software development, for modeling risk, security requirements, and so on. The integration of models expressed in i* variants poses interoperability problems. This issue has been faced at different levels, e.g. through unified metamodels, or with an interchange format to depict i* models. In our research work, we propose a practical approach to tackle the interoperability problems, exploiting ontology-based techniques. In a previous work, we presented an ontological representation of the i* metamodel and a tool to automatically transform an i* model to an instance of this ontology. In this position paper, we describe a methodology to integrate i* variants through the ontological representation of the i* metamodel, and a mechanism to support the understanding of models expressed in those variants. As example of application, we present the integration of i*, Tropos and Service-oriented i* by using ontologies and a tool to automatically transform models expressed with those variants in terms of that ontology.
The i* Framework [
The main objective of our research work [
The first specific objective has been addressed in [
Our scientific contributions are related to the accomplishment of the specific objectives 2, 3 and 4 of our research work. Therefore, in this section we present our proposed methodology, which describes how to use ontologies to achieve the integration of i* variants. Moreover, we present the application of the methodology to i*, Tropos and Service-oriented i*.
Our proposed integration methodology provides the guidelines to integrate
constructs of several i* variants into an ontology which extends OntoiStar [
In this phase the ontology for a specific i* variant is generated. The resultant ontology is based on OntoiStar. Therefore, additional constructs of a specific i* variant are added into OntoiStar. This phase may be performed several times when more i* variants need to be integrated. It consist of four steps: I. Identify. The first step corresponds with the identification of the additional constructs of the i* variant which are not part of the ontology OntoiStar. II. Categorize. The second step corresponds to the categorization of the additional constructs identified in the first step. Four categories allied with the elements of metamodels have been defined. A construct is categorized as: Concept, when it corresponds to a representation of something in the real world. Relationship, when it corresponds to a relationship of two or more concepts. Attribute, when it is used to define a property or characteristic of a concept. It may also refer to or set the specific value for a given instance of such. Attribute value, when it corresponds to those values that belong to an additional construct which has been categorized as attribute.
III. Transform. The third step corresponds to the transformation of the additional constructs into ontological constructs, i.e. classes, properties and axioms in OWL. Two sets of transformation rules have been proposed according with the i* variant specification. One for those constructs identified from a metamodel described in the UML language, and other for constructs identified from a textual description (Table 1).
IV. Classify. In [
Actor Relationship, if the OWL class describes a new type of actor relationship. Dependency, if the OWL class describes a new dependency relationship. The dependency basic structure has been already defined in OntoiStar. Boundary, if the OWL class describes a new type of boundary.
Intentional Element, if the OWL class describes a new type of intentional element.
Intentional Element Relationship, if the OWL class describes a new type of intentional element relationship.
In this phase, the ontologies of the i* variants (generated in phase 1) are integrated by means of an iterative merging process, obtaining as a result, the ontology OntoiStar+. The merging process consist of applying a merging function to the i* variant ontologies, two at a time, till obtain the ontology with all the desired i* variants. It is applied first to two i* variant ontologies, then, the resultant ontology is merged with another i* variant ontology, and so on. See for instance Fig. 1. The merging function consist of bringing together all the constructs of two ontologies, taking into account that duplicated constructs are only considered one time in the final merged ontology. The merging function has been implemented in the tool described in 3.3. With this approach a user can select the ontologies to merge according to the i* variants the user works with.
We have applied the integration methodology to i*, Tropos and Service-oriented i*. The integration results are presented in Table 2 (attributes were omitted due to space). The first two columns describe constructs already included into OntoiStar. The next three columns contain the additional constructs of each variant (N.A. indicates there is no additional constructs). First, we performed Phase 1, therefore, additional constructs of each variant were identified, categorized, transformed and classified in order to obtain the ontology for each variant. For instance, in Service-oriented i*, the construct “Service” was identified, then, it was categorized as concept and transformed as an OWL class. Finally, it was classified as an Intentional Element, therefore, placed as a sub class of the Intentional Element class in the OntoiStar taxonomy. Then, Phase 2 was carried out by merging the ontology for i*, the ontology for Tropos and the ontology for Service-oriented i*, obtaining as a result the ontology OntoiStar+ with the three variants integrated. The integration process is represented in Fig. 1. The rIenletaleetnimotnieonsnhtai pl CDMoenecttaorinimobsnupE,tonisodin-, last column of Table 2 contains the constructs included in the resultant ontology OntoiStar+.
In this section, we present TAGOOn (Tool for the Automatic Generation of
Organizational Ontologies), an extension of the one in [
In this paper, we presented a further step in the achievement of interoperability
of i* variants, which extends a previous work were an ontological representation
of the i* metamodel was proposed [
The main steps we intend to address in our future work can be summarized as follow: a) To take into account the semantic of the i* variants during the creation of ontologies; b) To propose inference rules that enable the redefinition and adjustment of i* based models according with the semantic and differences of i* variants; c) To extend TAGOOn to carry out the automatic transformation of i* based models from one i* variant to other i* variant.