Currently, i* is one of the most well founded organizational modelling techniques. Its main feature is the expressibility to represent intentional social relations among stakeholders. In i* models, each modeling component is described explicitly through text labels. However, the process of labeling model elements is usually an activity which is not rigorous and not well documented for designers. Performing the labeling with freedom and subjectivity often results in unclear labels that are not helpful for interoperability and the understanding of the model semantics. In this paper, we deal with this problems by extending i* models with ontologies. Taking advantage of an ontological de nition of concepts and well-de ned relationships, we improve the unambiguous interpretation of labels and thus interoperability, reuse and machine-readability of a model. A guided process and tool support for the integration of i* models with an ontology are described in this paper.
The i* framework is a goal-oriented and agent-oriented modeling framework. It
provides the needed infrastructure to model concepts such as actors, roles and
agents, and to reason about them. Nowadays, many research projects exist that
use the i* in di erent applications domain on early requirements engineering,
business process design and system requirements [
However, the absence of guidelines or good-practices to label business elements usually leads to the subjectivity, resulting in ambiguous labels that make the models di cult to understand for both the analysts and the target audience. 2 Furthermore, the amount of information that can be encoded in a human readable label is necessarily limited. Thus, the interpretation of an organizational model can become inevitably complex. Moreover, the machine-readability of a model remains quite limited.
In the remainder of this paper we present an approach to extend i* models
with semantic annotations taken of general or domain ontologies. This allows us
the standardization of concepts, clarifying the labels that describe an element
and also it permits to improve the analysis of existing models. Section 3.1
describes a set of semantic suggestions to guide the process of model annotation
based on domain and general ontologies. Section 3.2 describes a tool-supported
approach for combining the i* model (in its ontological representation [
As rst objective of this work, we propose the extension of i* models with concepts taken from an ontology in order to address the above-mentioned ambiguity issues that emerge from the labeling of organizational models.The enrichment of models with well-de ned concepts allows us to clarify the labels that describe an element to improve the analysis of existing models. Starting from this, as a second objective, we try to explore the possibilities given by a (tool-supported) ontological representation of the annotated i* model joined with an ontology.
To address our rst objective, we extend i* models with annotations from a
well-de ned ontology. The extension process consists of three steps(i) we describe
a set of suggestions applicable to ontologies, these suggestions are the key to
annotate the i* models; (ii) we propose an extension of the iStarML [
To address our second objective, we translate the i* model into an
ontology [
The core of our contribution is the extension of i* models with semantic annotations taken of ontologies. The top of Fig. 1 presents the proposed approach and the bottom side presents an example of this approach.
This process is composed by three main steps.
Step 1. Semantic annotation suggestions. To annotate model elements with
concepts from an ontology, we rst de ne guidelines, in the form of a set of
semantic annotation suggestions. An ontology represents knowledge as concepts
and relationships between them. Speci cally, general ontologies such as DOLCE
and UFO describes general concepts like space, time, matter, event, etc. [
To develop the suggestions, a semantic analysis of the primitives in i* and its variants was carried out. We analysed and compared the de nitions among primitives of the same type (e.g., goal ), in order to identify the di erences and similarities among them, obtaining a single de nition for each primitive. Next, analysing the structure of general and domain ontologies we tried to nd matches between the obtained de nition of the primitives and the concepts and relationships of the ontologies, with the goal to formally establish the relation of each obtained de nition with ontology concepts. We provide a set of general semantic annotation suggestions (Table 1) and a set of speci c semantic annotation suggestions (Table 2). The general suggestions are applicable to any general ontologies, while the speci c suggestions are applicable to the OntoSem ontology and its extensions.
We illustrate the annotation process with two shorts examples. Let's assume that a goal element labeled \Get credit" is annotated using a domain ontology on nancial operations. Using the general suggestions for elements of type \goal"(Table 1), the goal \Get credit" can be annotated with \Financing" and \Credit request". Now, let's assume that a goal element is labeled \Present card for transaction" (bottom side of Fig. 1) using OntoSem and the 4 suggestions for elements of type \goal" (Table 2), where \ME" means Model Element, \ A!B " means can be annotated and \SC" means SuperConcept, this element could be annotated with \Negotiate transaction", \Authenticate" and \Identify". The idea is to rst follow the semantic suggestions, then to go indeep in the selected ontology and to nd out the most appropriate concept for each model element, in a manual process. This approach could be used with di erent ontologies independently of the model domain thanks to the semantic annotation suggestions. Primitive Suggestion
Actor An actor (including the actor types) should be mapped into domain concepts that describe an organization, agent, tangible entity, or intangible entity.
Goal A goal should be mapped into domain concepts that describe a clear and precise condition, interest or desire.
Softgoal A softgoal should be mapped into domain concepts that describe an interest or desires not clear-cut satisfaction criteria.
Task A task should be mapped into domain concepts that describe a concrete action or activity. Resource A resource should be mapped into domain concepts that represent a physical object or informational entity.
Step 2. Extension of iStarML. To provide a format for interoperability between iStar modelling tools and dialects, we extend the iStarML model interchange format, adding an XML attribute called sannotation. This attribute stores the semantic annotations for each model element, with the syntax sannotation =\concept1 concept2...conceptn".
Step 3. Extension of an existing plug-in for jUCMNAv. To use semantic annotation in practice, we use the popular iStar modeller jUCMNav, adding the annotations with the symbol \@ ". We extended an existing plug-in for jUCMNAv to generate the iStarML les containing the annotations.
We provide a tool-supported process for joining annotated i* models with
general or domain ontologies. First, an organizational ontology is created from an
i* model by following the approach presented by Najera et al. [
We enhanced the TAGOOn tool (Tool for the Automatic Generation of
Organizational Ontologies ) provided by [
Taking an i* model (MiStar), TAGOOn+ creates its ontological representation (OM ) by using the functionalities provided by the original TAGOOn tool, instantiating each model element to an individual in OM . Next, it analyzes the selected ontology (OD) to obtain the hierarchy and description of each concept. OM and OD are joined, determining if an individual element was annotated with concepts (classes of the selected ontology). If an element is related with one or more concepts, a relation of type is-a is created between the concept in OD and the instantiation of the model element in OM . The obtained combined ontology (OI ) thus integrates the domain knowledge of OD and the organizational and intentional perspective provided by MiStar. For instance, a model element ME is annotated with the concept C1 from an ontology O1. After the integration, the ME is an individual element IE that presents a relation of type is-a with C1. The resulting ontology in OWL format can be visualized and edited with the popular Protege ontology editor. In Figure 1 (bottom right side) fragments of OntoSem and i* models are visualized, both integrated into a single ontology.
Depending on the properties of the domain or general ontology used, Protege reasoning engines o er consistency checking, inference and classi cation. Moreover, TAGOOn+ can create a text document that describes each element of the model with its semantic annotation and its description taken from the selected ontology. The documentation can be useful for achieving a better understanding of the i* model, for sharing and reuse.
Our approach has been used to annotate models that described a farm
systems, a generic card-based payment system, and the processes to register
students at a postgraduate institution [
In this paper we presented an overview of our approach to extend i* models with concepts taken from ontologies. A set of semantic suggestions guide the annotation of i* models were presented. A model with semantic annotations from an ontology is clear for humans and accessible to machines. Moreover, it can help improving the labeling quality in particular for models that evolve for a long time or that need to be read or edited by various analysts and stakeholders with a di erent background. Joining the model (in its ontological representation) with a generic or domain ontology, depending on the properties and axioms de ned in the ontology adopted, the obtained joint ontology could be checked for consistency and completeness. We provide tool support for the complete process and tested it with available iStar models and ontologies obtained from web repositories. The bene ts of extending an i* model with annotations would be, rst, to facilitate the analysis and understanding of a model proving a clear model supported for domain concepts and second, as our approach is based on iStarML, we permit the interoperability among i* variants through domain concepts.
At the present time, we are focusing on extending i* by describing its elements with generic concepts. As a future work we attempt to use natural language processing techniques for the annotation of each model element. In this way, semi-automatic suggestions will be provided to annotate the model. Moreover, we consider that a concept that integrates di erent model elements could represent new business services to the organization. These new functionalities can be useful to delineate new business services, and to improve the understandability and expressiveness of a model, thus giving a necessary condition for model reuse.
Finally, an empirical study and the modelling of real-world systems would be needed to give practical and statistical evidence to the e ciency of the approach.