Usually, enterprise models consider different aspects and include different abstraction levels of enterprises. The application of ontologies as conceptual bases that can clarify relations within and between these abstraction levels is believed to be helpful. This paper investigates the use of ontologies for formalizing enterprise modelling languages and enriching their semantics. The aim is to transform enterprise models into ontologies based on a mapping of the enterprise models' meta-model into a semantically corresponding ontology. The ontology representation then is used to check logical consistency and to infer new facts regarding the implications of the model beyond what would be possible with a visual modelling language. In order to check feasibility and pertinence of our approach, we selected the goal modelling part of the 4EM method. This paper provides (1) a formal OWL representation of the 4EM goals meta-model; (2) a systematization of transitive goal properties; (3) a set of SWRL rules expressing these transitivity; and (4) an analysis of exemplary goals model instances.
In general terms, enterprise modelling is addressing the systematic analysis and
modelling of processes, organization structures, products, IT-systems or other
perspectives relevant for the modelling purpose [
This paper investigates the use of ontologies for formalizing enterprise modelling
languages and enriching their semantics. The focus in this context is on visual
languages which have the advantage to be better understandable by non-experts in
enterprises but which in most cases lack operational semantics (see [
The ontology representation then is used to check logical consistency and to infer new facts regarding the implications of the model beyond what would be possible with a visual modelling language. In order to check feasibility and pertinence of our approach, we selected one modelling language (4EM; see section 2) and focused within 4EM on the goal modelling part.
This paper provides (1) a formal OWL representation of the 4EM Goals metamodel; (2) a systematization of transitive goal properties; (3) a set of SWRL rules expressing this transitivity; and (4) an analysis of exemplary goals model instances.
The remainder of the paper is structured as follows: Section 2 describes the
construction of an ontology representing the 4EM meta-model for goal modelling.
Section 3 shows how this ontology can be enriched by adding transitivity rules.
Section 4 provides a validation of the formalized meta-model by instantiating an
example model coming with the 4EM specification in [
Experience reports on Enterprise Modelling indicate both, the usefulness of ontology
representations [
From the existing EM methods, the „For Enterprise Modelling (4EM)” [
The Goals Model focuses on describing the goals of the enterprise. This model captures what the enterprise and its employees want to achieve, or to avoid. Goals Models usually clarify questions, such as: ─ Are there conflict/support relationships between goals? ─ Are there constraints/problems that hinder the achievement of a goal? ─ What sub-goals have to be achieved in order to achieve a goal? ─ What generally hinders/supports the achievement of a goal?
These so called Competency Questions (QC) can be used as a requirements
specification for an ontology on the domain of enterprise goals [
Especially in complex models visual analysis of these aspects is error prone. If a sub-goal is in a conflict or underlies some constraints, these circumstances should also be considered at top-level. Furthermore, inherent inconsistencies like supporting toplevel goals having conflicting sub-goals need attention. Ontology-based reasoning provides a tool to assess these issues stemming from transitive relationships in goal modelling.
In the following, the ontological representation of the 4EM Goals meta-model will
be constructed according to the 4EM method description in [5, pp. 87-101]. First the
taxonomy of goals model component types (classes) is constructed. In a second step,
the construction of binary and n-ary relationship types follows. Relationship
transitivity is discussed separately in section 3 because it is not specified in [
The model component types are represented as classes in OWL. All goals model component types are represented as specializations of the abstract class GM_ModellingComponent. The Goal class represents goals or objectives respectively. The 4EM method describes priority and criticality as optional attributes for goals. These have not been considered in the meta-model so far. This is kept for later work. Problems symbolize environmental circumstances that hinder the achievement of goals. Problems can be described more specifically as weaknesses (internal factors) and threats (external factors. Problems are represented in OWL with the Goal class and its sub-classes Threat and Weakness. A cause expresses explanations or reasons for problems (Cause class). Apart from causes, constraints (Constraint class) express business restrictions, laws or external policies that affect components of the goals model. The last component type are opportunities (Opportunity class) which symbolize resources supporting the achievement of certain goals. 2.2
Relationship types are represented as object properties in OWL. Object properties are directed binary relationships. Further semantics can be added to object properties by defining characteristics like transitivity and relations to other object properties, including specialization/generalization.
The 4EM goals model describes four binary relationship types. First, the supports relationship shows that fulfilling one goal also supports the achievement of another. Furthermore, the relationship is used to relate opportunities to goals. The contradicts relationship in contrast shows that the achievement of one goal is in conflict with another. This relationship is considered to be symmetric. Hence, if goal A contradicts goal B also goal B contradicts goal A. The hinders relationship is less strict. It can be used between model components to show negative influences. This relationship is not considered symmetric but can also be used to link goals. The last binary relationship is the causes relationship. It is used to link causes to problems.
Experience from ontology engineering shows that inverse relationships should be included in an ontology in order to fully specify concept relationships. For example, a problem can be linked to one of its causes by a caused_by relationship. These inverse relationships are automatically added to instances by OWL reasoning if defined in the meta-model. Table 1 shows the specification of the binary relationships. Object Property supports contradicts hinders causes Two additional abstract classes have been added. Supporter for goals model element types that can support the achievement of a goal (sub-classes Goal and Opportunity) and Hinderer for element types that can have a negative influence on the achievement of a goal (sub-classes Goal, Problem, and Constraint). The supports relationship is considered to be transitive. Hence, if A supports B and B supports C, A also supports C. Similar assumptions cannot be made for the other binary relationships. 2.3
N-ary relationships define semantics of goal decomposition in the 4EM goals model.
The AND-relationship decomposes a top-goal into a set of sub-goals that have to be
fulfilled each in order to achieve the top-goal. The OR-relationship defines a set of
sub-goals where it is sufficient to fulfill one of the alternatives. Finally the
AND/ORrelationship needs a combination of some of the sub-goals to be fulfilled. N-ary
relationships are not directly supported in OWL. Logical Ontology Design Patterns
can be used in order to model cases where the ontology language does not provide
appropriate constructs [
The rows contain the object properties to be propagated. Referring to the discussion in section 2, these include the relationships originally defined by the 4EM method and their inverse properties as well. For example, a goal can hinder another goal and can be hindered by some Hinderer as well.
A first decision made for transitivity specification is the exclusion of property propagation via goal conflicts and hinders relationships. For example, if goal B hinders goal A and goal B is hindered by goal C no assumption can be made that goal C supports goal A (double negation). The same applies for the contradiction relationship. Influences like constraints and problems that pose difficulties for the achievement of a goal may not influence the achievement of another goal at all or may also have a negative influence on it. Furthermore, goals are desired future states. Conflicts between goals need to be solved by a decision in favor of one of the goals or by relating the degree of goal fulfillment. The focus should be on the goals not on relating the context of one goal to the other.
The situation is different for supports relationships between goals. The semantics
of these relationships means that a sub-goal is a more specified part of the top-goal
(cf. [
Propagating hinders, supports and contradicts via supports relationships is not considered. In contrast to the AND composition, the sub-goal is not required to be fulfilled in order to achieve the supported top-goal. For example, if goal A supports goal B and goal A hinders goal C, it cannot be concluded that the fulfillment of goal A also hinders goal C.
After clarifying which object property propagation semantics should be supported, a formalization of these semantics is required. Generally, there are two possibilities to add such object property related semantics for inference mechanisms. First, the OWL language can be used. Here, object property axioms provide means to infer object property assertions (relationships between instances) based on existing object property assertions. Second, a rule language like SWRL can be used. Here, new facts are inferred based on a test of freely defined OWL statements against the ontology. If the body of a rule is found to be true its head is considered true as well and a new fact can be added to the ontology.
Transitivity of the supports relationship has already been defined in section 2 and can be expressed in OWL. However, property chains as introduced in section 2 are not fully supported by current OWL reasoning tools (Hermit 1.3.8). Thus, only SWRL rules are used to address the transitivity along property chains and n-ary relations. Table 4 shows the resulting formalization of the transitivity rules discussed in section 3.1. An ontology containing the instances of the example from section 4 can be found here: http://win.informatik.uni-rostock.de/uploads/media/4EM_GM.owl hindered by supported by contradicts AND composed by OR composed by AND/OR composed by
AND composed in subGoalComposedIn(?SubGoal,?Comp), compositionTopGoal(?Comp,?TopGoal) -> hindered by supports(?SubGoal,?TopGoal) hinders(?c,?SubGoal),supports(?SubGoal,?TopGoal) -> hinders(?c,?TopGoal) supported by subGoalComposedIn(?SubGoal,?Comp), compositionTopGoal(?Comp,?TopGoal) -> supports(?SubGoal,?TopGoal) supports is defined transitive
ANDGoals(?ANDComp), compositionSubGoal(?ANDComp, ?SubGoal), contradicts compositionTopGoal(?ANDComp,?TopGoal),contradicts(?c,?SubGoal) -> contradicts(?c, ?TopGoal)
ANDGoals(?ANDComp), compositionSubGoal(?ANDComp, ?SubGoal), cAoNmDposed by c?AoSNmuDpbGoSosuaibltGsio(oa?nlAT)N,DopSGcuoobamCpolom(sp?i)ANt,iDCoconoTmmpop,poGs?oiTatolpi(Go?onASaNulD)bS,GuobaClo(m?pA,N?DSSuubbGCooamlp), -> compositionSubGoal(?ANDComp, ?SubSubGoal) OR composed subGoalComposedIn(?SubGoal,?Comp), by compositionTopGoal(?Comp,?TopGoal) -> AND/OR supports(?SubGoal,?TopGoal) composed by hinders supports contradicts supports is defined transitive ANDGoals(?ANDComp), compositionSubGoal(?ANDComp, ?SubGoal), compositionTopGoal(?ANDComp, ?TopGoal), hinders(?SubGoal, ?c) -> hinders(?TopGoal, ?c) ANDGoals(?ANDComp), compositionSubGoal(?ANDComp, ?SubGoal), compositionTopGoal(?ANDComp, ?TopGoal), supports(?SubGoal,?c) -> supports(?TopGoal, ?c) contradicts is defined symmetric 4
In order to assess the applicability of the ontology to 4EM Goals models and the
benefits of OWL reasoning, we have adopted the exemplary A4Y case from [
It was possible to instantiate the complete model using the ontology. Furthermore, inference with the Hermit 1.3.8 reasoner added new facts to the model. No logical errors have been found. Regarding the relationship between Goal2 and Goal 3 it was inferred that Goal 2 hinders Goal 3 because Goal 10 is necessary to achieve Goal 2 and hinders Goal 3 at the same time (see Fig. 3). Thus, even if those hidden relationships are not modelled directly they reveal by reasoning using the proposed ontology schema and rules. Furthermore, the complete context is constructed automatically for a goal. All hindering and supporting influences are assigned to the goals for detailed analysis. Thus the Competency Questions formulated in section 2 can be answered by the ontology. The ontology allows for inferring hidden contradictions and hinders relations as described for the case of Goal 2 and Goal 3. These could be missed when relying on visual analysis only. Additionally, ontology based queries can be performed for further analysis. For example, goals that have hinders and supports relationships to each other at the same time need special attention and can be identified (G2 supports and hinders G1 in Fig. 3). Reasons may be conflicting sub-goals.
Based on the example of 4EM goal modelling, this paper investigated the possibility to transform meta-models of existing enterprise modelling languages into ontologies. The purpose of this transformation was to further specify the relations between focal areas and the constructs within a focal area, to check logical consistency, and to infer new facts regarding the implications of the model beyond what would be possible with a visual modelling language.
Our work showed that the developed ontology is applicable and the implemented reasoning provides support for analysis of the goal model.
Future work will have to investigate the implications of an ontology-based formalization for 4EM and the transferability of results to other enterprise modelling languages. In order to understand the implications for 4EM we started to capture the complete meta-model of 4EM in an ontology, i.e. to extend the goal modelling ontology to all perspectives of 4EM. This overall 4EM ontology will have to be used to check inconsistencies and clarification needs in 4EM. We expect that more transitivity rules and reverse relationships will have to be added. Regarding transferability to other enterprise modelling languages, we do not expect general problems as long as the language in question does not define operational semantics.