In this chapter, we has investigated the diversity of the modeling constructs defined in a number of existing business process modeling languages. The modeling constructs have been categorized according to the process perspectives. The categories illuminates the possibility to map the modeling constructs between different modeling languages in each process perspective. We have also surveyed a number of different process ontologies. Through the survey, we have compared the ontological representations of process perspectives in different process ontologies. The comparison results provide some principles of ontological representations of a process ontology which we can apply in our process ontology proposal for the semantic annotation purpose.
3.6. SUMMARY 59 In the goal modeling survey, we have found some overlapping modeling constructs in goal modeling and process modeling, such as actor or agent, resource or object, task or operation and constraint. That discloses the underlying relationships between goal models and process models. Connecting relationships in EEML goal modeling provide more explicit view of such links between goals and processes.
A list of requirements have been identified based on the survey of the annotation tools and methods. Although they are not tailored to the semantic annotation of process models, they provide a good baseline — what a semantic annotation tool should look like and how to develop an ontology-based semantic annotation approach.
Part II
Design and Application
Chapter 4
Semantic Annotation Framework
for Process Models
As we have discussed previously, process properties of a business process model are represented by meta-model elements/modeling constructs of a modeling language, and model solutions are represented by model elements composed of the modeling con- structs. Common expressions of process properties and context references make the heterogeneous process models reconciled and comparable. The enhanced semantic in- teroperability can facilitate the management and reuse of those process models. In this chapter, a basic semantic annotation framework [94] [92] [93] for process models is presented. The framework provides a common semantic schema to express process properties and reference the heterogeneous solutions of business process models to a consensual representation of systems’ context.
Before we present our framework, we discuss the theoretical basis of this work, and then we elaborate the structure and components of the framework. Under such a framework, a general process ontology is introduced and used as the reference ontology to annotate process properties. A set of semantic mapping rules are defined as the annotation method, and a semantic annotation model is introduced as the annotation schema.
4.1
Theoretical Basis
The theoretical basis underlying our proposal is Ogden & Richard’s semiotic triangle [121], which has been mentioned in section 2.2.1. We use the semiotic triangle to analyze semantic heterogeneity of process models. Ontology plays a major role in our semantic annotation approach, which is also fitted into the semiotic triangle. Thus, in this section we discuss about the relationship between the process models and ontology in the context of semantic interoperability based on the semiotic triangle.
4.1.1 Semiotic triangle
The semiotic triangle is applied in both information modeling and philosophy disci- plines. The theory describes relationships between reference, sign and concept (refer to
64 CHAPTER 4. SEMANTIC ANNOTATION FRAMEWORK section 2.2.1). Based on the semiotic triangle we can build the relationships between model, meta-model, modeling language and ontology as shown in Figure 4.1. In the tri- angle, a model is a conceptualization of referents and it is represented as a set of model denotations in a certain modeling language. Model denotations are signs which signify concepts in the model. The model is an instance of a meta-model, and the meta-model defines a modeling language. A concept is a mental perception which is in a human’s mind. One concept referring to a referent can be represented differently. On the other hand, representations of different concepts may look similar in two models. The dif- ferences are results of the way of conceptualization of modeling or representations of model denotations defined in a modeling language. The semantic heterogeneity exist- ing in different models can therefore be distinguished at the model and the modeling language level (we also call it the meta-model level).
Figure 4.1: Relationships between ontology, model, meta-model and modeling language in the semiotic triangle
In order for a machine to understand the heterogeneous semantics in the models (e.g. various signs of referents referring to the same concepts or synonym signs of referents referring to different concepts), a common understanding of concepts has to be formalized in a machine-interpretable way. An ontology is created for this purpose. In the semiotic triangle, the semantics of concepts are formally in a standard i.e. they are represented as an ontology. Ontologies aid the sharing of knowledge on the basis of the assumption that there is a single reality and the sharing is a matter of aligning the way different people or systems think about it [70]. Therefore, in the semiotic triangle
4.1. THEORETICAL BASIS 65 concepts can be conceptualized as an ontology in a consensual way.
A meta-model is also a model – a model of the modeling language. For a meta- model, a modeling referent is a meta-model element or modeling construct, and a modeling sign is the notation of a meta-model element. A modeling sign standing for a meta-model element is used to represent a modeling concept in a certain modeling domain. Semantic heterogeneity problems will still occur on the meta-model level pro- vided that a meta-model is a model. We assume that a modeling ontology can provide the common conceptualization of modeling concepts referring to the same modeling ref- erents. According to Leppänen’s OntoFrame [87], a meta-model can be adapted from a modeling ontology. The semantic heterogeneity of modeling languages can there- fore be reconciled through the modeling ontology. Based on this theory, we annotate the meta-model of a modeling language with the modeling ontology in this research. The relationship between modeling language, meta-model and modeling ontology is displayed in Figure 4.2.
Figure 4.2: Relationships between modeling ontology, meta-model and modeling language in the semiotic triangle
4.1.2 Semiotic triangle for process modeling
In the context of our research, process meta-model, process modeling language, process model, process model denotation and process ontology are specified in the semiotic triangle (Figure 4.3). We also include model levels in the figure to explicate the positions of those modeling concepts. The model levels are adapted from the model level ontology in [87] to show the different levels of process models. Process meta-model and process model are both models. Process meta-models are categorized at the meta level, and process models are at the type level. In this research, we focus on process models at the type level including their meta models at the meta level. Process models at the type level are the resources of process knowledge in our context.
66 CHAPTER 4. SEMANTIC ANNOTATION FRAMEWORK
Figure 4.3: Relationships between model level, process ontology, process model, process meta- model and process modeling language (adapted from model level ontology in [87])