The purpose of model annotation is to annotate model (contents) with domain ontolo- gies. Models (contents) are instances of the meta-model and those instances usually describe certain domains. The representations of domains are often various due to diverse uses of terminology and conceptualization, resulting in semantic heterogeneity of model contents. Domain ontologies are agreed as standard representations and se- mantic definitions of domain concepts by annotation users. Semantic heterogeneity of model (contents) can be reconciled by referencing ontological concepts represented in domain ontologies. In model annotation, the annotation method is building semantic mappings or relationships between model contents and domain ontologies.
The mapping method for model annotation
Different mapping strategies can be used between concepts in the model and the do- main specific ontology. They can be simple rules applied in meta-model annotation – by referring specific model contents in modeling constructs to corresponding domain concepts. More complicated mappings can be defined through refined semantic rela- tionships between concepts used in models and concepts defined in a domain ontology. Simple reference. If a simple mapping by reference is applied, it assumes that almost all concepts in the model have equal or approximately equal concepts in the
4.5. MODEL ANNOTATION 75 ontology. The semantic relationship of mapping can be defined as one type – refers_to. We have adopted such a mapping strategy in our method for meta-model annotation to build the correspondences of concepts between modeling languages and GPO. In the model annotation, users can choose this strategy provided the concepts in the models are very close to the concepts in the domain specific ontology. The strategy of simple reference is easy to apply and also makes it relatively simple for the machine to infer the mapping relationships without complicated algorithms.
Refined semantic relationships. Concepts used in process models are variously defined initially for different projects. Therefore, it might be difficult to find equally defined concepts in the domain specific ontology for process models. However, since they are still within one domain, there must be some semantic relationships between the concepts in models and in ontologies. In order to represent the semantic relationships precisely, we define some refined semantic relationships to link the concepts between models and ontologies for the model annotation.
The model annotation can be accomplished with the help of meta-model annotation, because the models related to domain information are usually artifacts, actor-roles, activities and exceptions. Artifacts and actor-roles are usually static concepts in the domain ontology. Activities and exceptions are usually related to the task concepts in the domain ontology.
Semantic relationships and the corresponding annotation denotations generally used for the model annotation are listed in Table 4.2. Both synonym and polysemy relation- ships are symmetrical. Note that same_as and different_from are not at the individual level like the OWL individuals relationships. Hypernym and hyponym are inverse rela- tionships. Usually concepts in the ontology are more general, while model concepts are relatively concrete for specific projects. Thus, kind_of is more often used than super- Concept_of when annotating model concepts with ontology concepts. We provide more human sense expressions of meronymic relationship for artifacts, actor-roles, activities and exceptions respectively. For artifacts, we use part_of, e.g. Engine is a part of Airplane; for actor-roles, we use member_of, e.g. Airline is a member of Air Alliance; for activities, we use phase_of, e.g. Flying is a phase of Traveling; for exceptions, we use partialEffect_of, e.g. Payment is cancelled is a partial effect of Booking has failed. The inverse relationship of meronym is holonymic relationship, which is seldom used in this framework because of the similar reason described for the hyponym relationship.
It is possible to represent some of the semantic relationships with OWL expres- sions, such as owl:equivalentClass for synonym, owl:disjointWith for polysemy, rdfs:subClassOffor hypernym and hyponym. Those OWL expressions can be used to displace the corresponding annotation notations for OWL inference. If an annotation model and a reference ontology are both represented in OWL, the inference can be made on the two OWL files. For example, a concept Cm in the annotation model is specified as same_as an ontology concept O1, and O1 is defined as rdfs:subClassOf another ontology concept O2 in the ontology. An inference result could be Cm is sub-Class of (kind_of) O2 too.
After the meta-model annotation, the model can be described as a GPO-annotated process model. Accordingly, the model annotation can be done directly in the GPO- annotated process model instead of the original model. Thereby, the model annotation notations will also be formalized in the process semantic annotation model in section
76 CHAPTER 4. SEMANTIC ANNOTATION FRAMEWORK
Table 4.2: Semantic relationships, corresponding annotation denotations, and OWL constructs
Semantic Relation- ship
Annotation Denotation OWL expression
Synonym alternative_name(terminology level)
same_as(concept level) owl:equivalentClass
Polysemy different_from owl:disjoinWith
Hypernym kind_of rdfs:subClassOf
Hyponym superConcept_of rdfs:subClassOf
Meronym part_of(Artifact) member_of(Actor-role) phase_of(Activity)
partialEffect_of(Exception) Holonym compositionConcept_of
Instance instance_of OWL Individual