Evaluation of modelling languages is important both to be able to select the most suitable languages according to the needs, and to improve existing languages. In this paper Business Process Modeling Notation (BPMN) is presented and analytically evaluated according to the Semiotic Quality Framework. BPMN is a functionally oriented language well suited for modeling within the domain of business processes, but probably also general processes not only within the business domain. The evaluation indicates that BPMN is easily learned for simple use, and Business Process Diagrams (BPDs) are relatively easy to understand. Tools may fairly easily map BPDs into the BPEL4WS format, but executable systems then require creation of Web Services representing the Activities in BPDs. An evaluation according to the BWW ontology is useful for finding ontological discrepancies, and the semiotic framework is useful for evaluating quality on a relatively general level. These methods thus complement each other.
Currently there exist a large number of different modelling languages. Many of them define overlapping concepts and usage areas, and consequently it is difficult for organizations to select the most appropriate language related to their needs. Traditionally the research community have focused more on creating new modelling languages than evaluating existing ones. However, evaluation of languages is important both to be able to select the most suitable ones, and to improve existing languages.
Conceptual modelling languages can be evaluated analytically and empirically. As
Gemino and Wand discuss in [
Section 2 presents BPMN including its notation. It provides some examples of Business Process Diagrams (BPDs), and relates BPMN to the BPEL4WS standard. Section 3 presents the Semiotic framework for evaluation, divided into parts for evaluating the quality of conceptual models and the quality of conceptual modelling languages. Section 4 presents an analytical evaluation of BPMN according to the semiotic framework. Section 5 gives a short summary of what the BWW ontology is, how it may be used to evaluate conceptual modelling languages, and discusses in what ways this can complement the evaluation according to the semiotic framework. Section 6 mentions related work, and Section 7 presents future work. The conclusion is given in section 8. 2 2.1
Business Process Modelling Notation (BPMN) is a notation aiming to be easily
understandable and usable to both business users and technical system developers [
BPMN defines Business Process Diagrams (BPD), which can be used to create graphical models especially useful for modelling business processes and their operations. It is based on a flowchart technique - models are networks of graphical objects (activities) with flow controls between them.
The BPMN 1.0 specification was developed by The Business Process Management Initiative (BPMI) and released in May 2004. BPMN is based on the revision of other notations and methodologies, especially UML Activity Diagram, UML EDOC Business Process, IDEF, ebXML BPSS, Activity-Decision Flow (ADF) Diagram, RosettaNet, LOVeM and Event-Process Chains (EPCs). 2.2
The graphical elements that are defined by BPMN for use in BPDs are divided into a
small number of categories, so that they can be easily recognized even if a user is not
immediately familiar with a specific graphical element [
Flow Objects contain the three core elements used to create BPDs: Event (Start, Intermediate and End), Activity (atomic Task and compound Sub-Process) and Gateway (decision-making, forking and merging of paths).
Connecting Objects are used to connect Flow Objects to each other through arrows representing Sequence Flow, Message Flow and Association.
Swimlanes are used to group activities into separate categories for different functional capabilities or responsibilities (e.g. a role/participant). A Pool represents a Participant in a Process, and Pools can be divided into Lanes (e.g. between divisions in a company). Pools are used when a Process involves two or more business entities or participants. Activities within Pools must constitute self-contained Processes. Because of this, Sequence Flow may not cross from one Pool to another. Instead, Message Flow goes between Pools to indicate the communication between participants. See Section 2.4 for an example of this.
Artefacts may be added to a diagram where deemed appropriate. The following three Artefacts are defined: Data Object, Group and Annotation (to be used for comments and explanations).
For further introduction to BPMN, [
A metamodel is defined for BPMN [
To be better able to understand what BPDs are, two examples are shown here. Fig. 1
below shows a simple process using Flow Objects, Connecting Objects and Annotations.
Note how Sequence Flow is used in Fig. 2 only within the Pools, and Message Flow is
used for communication between the two Pools:
Business Process Execution Language for Web Services (BPEL4WS) is a standard for
specifying business process behaviour based on Web Services [
In [
The semiotic quality evaluation framework specifically distinguishes between goals
and means, meaning that they separate what you want to achieve from how you achieve it
[
The main concepts of the semiotic framework are Model, Modeling domain, Language
extension, Participant knowledge, Social actor interpretation and Technical actor
interpretation [
The seven different relationships represent different aspects of quality: Physical Quality regards the physical representation of the model and its externalization and internalization. Empirical Quality regards layout and a presentation that is easy to read and write without mistakes. Syntactic Quality is about the model being valid and complete with regards to the modeling language being used. Semantic Quality is about validity and completeness of the model in relation to the domain being modelled. Perceived Semantic Quality of a model is measured like semantic quality above, but in addition it depends on the actor's interpretation of the model and his/her knowledge of the domain. Pragmatic Quality regards the audience's comprehension of the model. Social Quality has the definition of actors having agreement (relative or absolute) about their interpretation, knowledge and model. 3.2
The semiotic framework for evaluating the quality of conceptual modelling languages is
based on the framework for quality of conceptual models [
Five aspects are identified for evaluating the quality of conceptual modelling languages: Domain appropriateness, participant language knowledge appropriateness, knowledge externalizability appropriateness, comprehensibility appropriateness and technical actor interpretation appropriateness.
The relationships in Fig. 3 below represent these five aspects of language quality:
- Domain Appropriateness. This deals with how suitable a language is for use within
different domains. If "there are no statements in the domain that cannot be expressed in
the language" [
The most central concept in BPMN is the process, which is built up from activities.
Because of this, the main perspective (as defined in [
But because of its functional perspective, BPMN has clear limitations to its domain
appropriateness. It is not well suited for expressing for example models in the
objectoriented domain. BPMN lacks concepts like class hierarchies. As stated in [
BPMN was created for the main purpose of modelling business processes, and is hence
well suited for modelling the business domain (e.g. B2B processes). However, the BPMN
1.0 Specification [
The graphical elements of BPMN are defined in a clear and concise way, to avoid confusions and ease the learning of the language. The language is also made to have similar notation to other languages like Flowcharts, UML Activity Diagrams, Event Process Chains, Petri Nets and Data Flow Diagrams (DFD). For example, a diamond shape is used both in BPMN, UML Activity Diagrams and Flowcharts to express a decision point, and both in BPMN and Activity Diagrams to express a merge. BPMN also have a striking resemblance to Activity Diagrams regarding the notational representation of events (small circles) and activities (rounded rectangles). In addition, the concept of Swim Lanes and its graphical representation are very similar in Activity Diagrams. Sequence flows are represented by arrows with solid lines and solid arrowheads in BPMN, Activity Diagrams, Flowcharts and Petri Nets. These similarities are helpful, at lest for IT professionals that are already familiar with the other languages. There are however also some graphical elements that are used differently in BPMN compared to other languages. For example, BPMN use a diamond shape also to represent forks or joins (of parallel activities), but in Activity Diagrams a thick horizontal line is used for this. Flowcharts use rounded rectangles to represent start- or end-states, and not to symbolize activities. These differences make it more difficult to learn BPMN.
It is a goal for BPMN that it should be understandable not only by IT professionals, but
also for business analysts and other non-technical people [
This area is highly dependable on the specific knowledge of the actors that are using the
language, and is therefore difficult to evaluate in a general way. We can however make
assumptions about the typical participants involved in the modelling process. BPMN
probably appeals the most to business users, since it was created especially for modelling
business processes. The term "business users" is however very broad and include a wide
variety of actors. If the actors want to model a process purely within the business domain,
BPMN has very good support for this. But if they desire to create models involving other
domains as well, this may be difficult (ref. subsection 4.1 above) and supplements to the
BPMN-models may be needed. Thus it may be hard for actors to externalize their relevant
knowledge using only Business Process Diagrams (BPDs) if that knowledge goes beyond
the domain of business processes. As already mentioned, the BPMN specification [
Comprehensibility of a conceptual modelling language can be divided into understanding
of language concepts and understanding of notation. Regarding notation, BPMN provides
a small number of notational categories so that the readers can easily recognize the basic
types of elements that constitute the diagrams [
BPMN support aggregation by allowing activities that are "collapsed" and contain subactivities. This helps to understand and get an overview over complex models. 4.5
BPDs are with a few exceptions easily mapped into the BPEL4WS-format. Guidelines for
doing this can be found in the BPMN 1.0 Specification [
BPEL4WS requires the use of WSDL and Web Services to be executable. Because of this, it is not so easy to do automated reasoning about processes that are not suitable to be implemented using a combination of Web Services.
Atomic Activities in BPDs are supposed to usually represent a Web Service. How to specifically implement these Web Services may be difficult to interpret, especially if the Activity is vaguely defined using only a short textual description. 4.6
The BPMN Metamodel (and its evaluation) is too complex for a detailed analysis in the scope of this paper. But its sheer size and complexity suggests that it might have a less than perfect pragmatic quality. This further strengthens our claim that normal business users without training will have difficulties understanding advanced business processes modelled using BPMN. 5
The Bunge-Wand-Weber model (BWW) is an ontological model of information systems,
and can be used to analyze and evaluate conceptual modeling languages [
Both methods are analytical in that they evaluate languages based on a theoretical framework. The results are lists of language features that correspond to the recommendations of the frameworks, and other features that have room for improvement.
The semiotic framework is well suited for evaluating quality on a relatively general level. The BWW ontology complements this by being more concrete as to evaluating and suggesting which concrete language constructs should be used. The BWW ontology looks at the conceptual basis of modelling languages, and can not be used to evaluate other aspects of a modelling language (as for example the diagram notation). Notation and other aspects can be evaluated using the semiotic framework, giving this technique a broad focus. The BWW ontology on the other hand has a narrower focus, but evaluations thus become more thorough. Evaluations using BWW are easily made more objective than when using the semiotic framework with its more general concepts. The semiotic framework and the BWW ontology complement each other as methods to analyze conceptual modelling languages.
UML is an example of a modelling language that has been evaluated both using the
BWW ontology and the semiotic framework. In [
The semiotic quality framework has been used to evaluate several conceptual modelling
languages. In [
The authors have not been able to find any other published papers evaluating BPMN.
However, some evaluations of BPEL4WS have been performed. This is relevant because
models created using BPMN in many cases can be mapped directly into a corresponding
model in BPEL4WS and vice versa. However, BPEL4WS-models are represented in
XML and have no graphical notation. In [
To further elaborate on the evaluation done in this paper, the quality of the documentation and tool support for BPMN should be analyzed using the semiotic framework.
Additional evaluation of BPMN should also be performed by comparing the BWW
ontology to the BPMN Metamodel. The comparison should look for construct overload,
redundancy, excess and deficit [
In addition to further analytical evaluation, empirical evaluation is needed to validate the results of the analytical investigations. Future work should also include comparative studies of BPMN and several other business process modelling languages. 8
BPMN is a functionally oriented language designed for easily modeling business processes, and is well suited for this domain. BPMN has limitations to usage within other domains (e.g. the object-oriented domain), but can be used to model general processes also outside the business domain. BPMN has a familiar and easy basic graphical notation, but also include complex and advanced features that probably require a fair amount of training for non-technical users to learn. BPDs have relatively good comprehensibility appropriateness due to categorization of the types of graphical elements and support for aggregation of Activities. Technical actors may fairly easily map BPDs into the BPEL4WS format, but creation of Web Services representing the Activities is required to make an executable system in this case.
It has been discussed how BPMN may be evaluated according to the BWW ontology, and in what ways this may supplement the evaluation according to the semiotic framework. An evaluation according to the BWW ontology is useful for finding ontological discrepancies, and the semiotic framework is useful for evaluating quality on a relatively general level. The semiotic framework and the BWW ontology complement each other as methods to analyze conceptual modelling languages.