Towards Hybrid Semantics of Enterprise Modeling Languages

Richard Braun and Werner Esswein

TU Dresden, Chair for Wirtschaftsinformatik, esp. System Development, 01062 Dresden, Germany

Keywords:

Enterprise Modeling, Meta Modeling, Formal Semantics, Material Semantics, Behavioral Semantics.

Abstract:

Enterprise Modeling Languages (EMLs) are generally perceived as conceptual modeling languages having

a formal syntax and informal semantics. The non-formality of semantics is mainly caused by the material-

ity of the addressed domain (enterprises and its related aspects) and the resulting personal interpretation of

syntactical constructs. However, EMLs may also explicitly deﬁne invariant interpretations in the sense of pos-

sible model executions or the deﬁnition of domain-speciﬁc restrictions. It is therefore promising to address

a possible amalgamation of material semantics and formal semantics in order to provide an integrated and

comprehensive semantic speciﬁcation of EMLs. This position paper introduces and motivates the topic by

systematizing and consolidating approaches from both ﬁelds and introduces a framework for so-called hybrid

semantics on the meta model layer. Further, the general relevance of semantics and semantic speciﬁcations in

EMLs is emphasized and prospective research challenges are proposed.

1 INTRODUCTION AND

MOTIVATION

1.1 Semantics in Enterprise Modeling

Languages

Enterprises are multifarious, heterogeneous socio-

technical organizations, whose components are in-

terrelated within a complex system on different ab-

straction levels (Vernadat, 2003). Enterprise Model-

ing (EM) aims to conceptualize, abstract and repre-

sent parts and aspects of enterprises by creating con-

ceptual models in order to foster communication be-

tween involved stakeholders and enable an integra-

tion of static, procedural and functional dimensions

(Lankhorst, 2009; Frank, 2013; Frank, 2014; Maes

and Poels, 2007).

Enterprise Modeling Languages (EMLs) serve

as languages for the construction of respectively

needed enterprise models. The range of EMLs can

be divided into integrated, domain-independent ap-

proaches like ArchiMate (Lankhorst et al., 2009),

ARIS (Scheer and N

uttgens, 2000) or MEMO (Frank,

2014), purpose-speciﬁc languages like BPMN (pro-

cesses (OMG, 2011)) or KAOS (risk management

(Heaven and Finkelstein, 2004)) and dedicated,

domain-speciﬁc approaches like CPmod (Burwitz

et al., 2013).

Most EMLs constitute as semi-formal modeling

languages (Wand and Weber, 2002), generally fea-

tured by a formal syntax and informal semantics

(Karagiannis and K

uhn, 2002). Informality of se-

mantics primarily results from the fact that EMLs ad-

dress the modeling of enterprises and related aspects,

which are usually not entirely formalizable (Pfeif-

fer and Gehlert, 2005), as they mainly refer to real-

world concepts (Lindland et al., 1994, p. 44). This

can be also referred as ontological semantics (Harza-

llah et al., 2012, p. 489). EMLs cover a practi-

cally unlimited set of concepts (Anaya et al., 2010,

p. 101). Semantics represents the meaning of a par-

ticular sign system (syntax) in regard of a semantic

domain (Overhage et al., 2012, p. 22).

Consequently, stakeholders have different concep-

tualizations and understandings of particularly ad-

dressed real-world things (van der Linden, 2015).

Also speciﬁc contexts (Bjekovic et al., 2014) or lex-

ical ambiguities (Delfmann et al., 2009) may deter-

mine different interpretations of one and the same

meta model construct. Besides this immanent fea-

ture of subject-dependent reference between enter-

prise models and semantic constructs (Rosemann

et al., 2004; Guizzardi, 2007; Opdahl and Henderson-

Sellers, 2002), invariant interpretation is explicitly de-

sired in some enterprise models for different reasons

(Bork and Fill, 2014, p. 3400). Besides, it could be

explicitly intended to limit the variance of interpret-

ing enterprise models in order to evolve and ensure

a common, coherent understanding within a particu-

412

Braun, R. and Esswein, W.

Towards Hybrid Semantics of Enterprise Modeling Languages.

DOI: 10.5220/0005812504120420

In Proceedings of the 4th International Conference on Model-Driven Engineering and Software Development (MODELSWARD 2016), pages 412-420

ISBN: 978-989-758-168-7

lar domain (Frank, 2013, pp. 13-14) - for instance, in

regulated and rather standardized domains like health-

care (Burwitz et al., 2013).

1.2 Research Objective and Research

Context

It is therefore advisable to focus the current state of

the art in regard of material semantics (Gehlert and

Esswein, 2006) and formal semantics (Bork and Fill,

2014) in order to investigate reasonable integrations

of these types. We therefore aim to provide a frame-

work for differing types of semantics based on their

characteristics. This is necessary, as the term seman-

tics itself (Thalheim, 2012, p. 7) has several seman-

tics within different research disciplines (Harel and

Rumpe, 2004, pp. 67-69). For instance, semantics

in formal modeling languages is understood as the

invariant transformation of one valid model state to

an other (Engels et al., 2000), while in semi-formal

languages a mapping to semantic concepts is meant

(Rosemann et al., 2004). This issue seems to be

closely related to the immanent multi-disciplinarity

in EM. The framework should further facilitate the

identiﬁcation of particular integration points and sup-

port the subsequent alignment and integration of the

doubtless existing description techniques for seman-

tics from different ﬁelds of research (van der Linden,

2015; Opdahl et al., 2012; H

offerer, 2007; Engels

et al., 2000).

The motivation of the stated research goal comes

from the rather poor semantic speciﬁcation of mod-

eling languages (Harel and Rumpe, 2004, pp. 67-

69) and EMLs (Santos et al., 2013, pp. 690, 706),

which hampers language comprehensibility (Guiz-

zardi, 2007; Wyssusek, 2006) and provokes mis-

understandings and semantic mismatches (van der

Linden, 2015). Further, syntax and semantics

(Natschl

ager, 2011b) or informal and formal aspects

are often amalgamated (Gehlert et al., 2005). Es-

pecially in OMG speciﬁcations, syntactical state-

ments and semantic explanations are often mixed and

distributed over hundreds of speciﬁcation document

pages (Hausmann et al., 2005, p. 19), causing misun-

derstandings and contradictions (Natschl

ager, 2011b).

This issue is even exacerbated due to the pure infor-

mality of semantic statements.

It is therefore necessary to elaborate and deﬁne a

generic and integrated speciﬁcation format for EML

semantics in order to provide possibly precise seman-

tics for meta model concepts, both in terms of variant

as well as potentially invariant semantics. This in turn

could facilitate semantic agreements in collaborative

EM (Rospocher et al., 2008; Renger et al., 2008).

While the formal syntax of EMLs is fundamen-

tal, we agree to Bjekovic et al. (2014) stating that se-

mantics and pragmatics are the actual drivers of EML

usage. Investigating material and formal semantics

in the EM domain is therefore also highly relevant in

terms of EML dissemination and revision, as seman-

tic comparisons might better justify the necessity for

EML dialects and language adaptations (Braun, 2015;

Bjekovic et al., 2013). Generally, general-purpose or

purpose-speciﬁc languages like BPMN or ArchiMate

provide generic semantics, which are often speci-

ﬁed or reﬁned for particular contexts (Hamann and

Gogolla, 2013, p. 501).

The remainder of this position paper is therefore

as follows. Section 2 introduces formal semantics in

detail, as the role of formality in EM is only little in-

vestigated so far. Section 3 then presents material se-

mantics. Section 4 outlines the aimed framework and

proposes several topics for following research. The

paper ends with a brief conclusion in Section 5.

2 FORMAL SEMANTICS

2.1 Relevance from the Perspective of

Enterprise Modeling

Formal semantics cover the unambiguous and

subject-independent usage of linguistic expressions,

which causes interpretive invariance (Guizzardi et al.,

2002). Formal semantics is relevant within enter-

prise modeling for basically two reasons: Automation

and analysis (Braun, 2015). Automation-oriented for-

mal semantics indirectly refers to the real-world do-

main in the sense of automatable process execution

in order to avoid failures and improve efﬁciency in

a proactive manner. Functional or process-oriented

languages are conducted for that purpose (Bork and

Fill, 2014). BPMN aims to support process execution

by textually describing their automatic interpretation

(OMG, 2011, p. 435).

While automation owns real-world references

(e.g., references to particular resources or machines),

in analysis-oriented formal semantics the transforma-

tion of models into other models or formalizations

is primal. For instance, deontic logic is applied to

BPMN (Natschl

ager, 2011b) and other authors intro-

duced speciﬁc operations on models for BPMN model

version management (Pietsch et al., 2014). However,

formal semantics is rarely discussed from the perspec-

tive of EM and there is a lack of specifying required

formal aspects in the context of EMLs (Bork and Fill,

2014). Below, generic approaches for both specifying

Towards Hybrid Semantics of Enterprise Modeling Languages

413

formal semantic domains as well as the formal seman-

tic mapping are hence introduced in order to prepare

a potential application for invariant parts of enterprise

modeling.

2.2 Formal Semantic Domain

The semantic domain of formal semantics does not

correspond to the real world, but rather to a con-

structed artiﬁcial world, which can be seen as solution

layer that aims to solve problems that are relevant for

real world problems but require further interpretation

(Gehlert, 2007; Pfeiffer and Gehlert, 2005). In de-

tail, a formal semantic domain often acts as kind of

a projection surface for solving consistency issues of

modeling languages, while the explicit real world ref-

erence is omitted (Engels et al., 2001, pp. 187-188).

Consequently, formal semantic domains are usu-

ally described by formal expressions (e.g., algebra

or graph-based languages), whose interpretations are

generally perceived as invariant (Tarski, 1936; Haus-

mann, 2003; Schobbens et al., 2006). Formal seman-

tics are mainly discussed in the context of the UML

(Engels et al., 2000), implicating a naturally strong

focus on design-oriented approaches within Software

Engineering (Hausmann et al., 2005, p. 5). Although

an explicit and integrated speciﬁcation of semantics

is deeply required (Maoz et al., 2010, p. 2), a lack of

formal speciﬁcations can be observed, e.g. for UML

(Soltenborn and Engels, 2009, p. 7).

Formal semantics can be divided into behavioral

formal semantics and static formal semantics (Engels

et al., 2000).

Behavioral formal semantics describe executable

things in the sense of transformations between states

in order to model token ﬂows or state transitions

(Hausmann et al., 2005, p. 25), aiming to enable auto-

matic interpretation of each model instance (Hamann

and Gogolla, 2013, p. 489). Consequently, abstract

state machines, Petri nets, automatons or ﬁnitie ac-

tions traces are leveraged as semantic domains for

behavioral semantics (Engels et al., 2000; Hausmann

et al., 2005; Maoz et al., 2010; Kossak et al., 2014).

There are further some works regarding execution

semantics of BPMN focusing workﬂow derivation

(Lam, 2012). By default, BPMN only provides rather

imprecise textual statements on this topic (OMG,

2011, p. 435).

Static formal semantics aims to represent valid

and consistent static states. Static semantics is rarely

speciﬁed in modeling languages and refers to deno-

tational approaches. It is hence more interesting to

limit a particular interpretation scope by introducing

additional constraints (e.g., regarding union subsets

in UML (Hamann and Gogolla, 2013)) or specifying

the technical domain that should be represented. For

instance, system models (Cengarle et al., 2014) are

used as formal domains of syntactical domains speci-

ﬁed with the MontiCore language (Maoz et al., 2010).

2.3 Formal Semantic Mapping

The actual formality of formal semantics results from

the invariant relation between syntactical constructs

and the semantics domain, which enables an a pri-

ori deﬁnition of the semantics (Messer, 1999, p. 99).

Formal semantics can hence be understood as replace-

ment of expressions of one language by expressions

of an other formal language causing syntactical map-

pings (Fill, 2015, p. 44). This actually leads to a con-

tinual shift of unambiguous interpretation (Gehlert,

2007, p. 37), constituting as rule-based chain of trans-

formations (Holten, 2003, p. 50), which typically

leads to a ﬁnal formulation of mathematic or algebraic

statements (Hausmann et al., 2005, p. 10).

Hence, formal semantic mapping represents the

speciﬁcation of two formal syntaxes; generally con-

stituting as operational or even axiomatic semantic

annotations (K

uhn, 2004, pp. 34-35). Surprisingly,

most mappings are only informally speciﬁed (Haus-

mann et al., 2005, pp. 25-27). Formal approaches

cover process algebra, linear equations, task graphs

or functions. In contrast to these textual approaches,

some researchers propose model-based approaches

for specifying relations between MOF-based meta

models (Hausmann, 2003).

Finally, the Dynamic Meta Modeling (DMM) ap-

proach has to be taken into account. DMM is a

generic approach for the integrated speciﬁcation of

formal semantics and mappings of behavior describ-

ing modeling languages, e.g., state charts or activity

diagrams (Engels et al., 2000). DMM only requires a

well-deﬁned meta model and can be implemented in

three major steps: First, the meta model of a language

is slightly extended. Then the behavior is formally

speciﬁed and a transformation into a typed graph has

to be conducted in order to realize the behavioral se-

mantics (Engels et al., 2000). Therefore, syntactical

expressions are coupled with pre conditions (events),

rules for transformation (action) and post conditions

(Soltenborn and Engels, 2009). If the pre conditions

for a syntactical construct are fulﬁlled then particular

rules are applied until the post conditions are satisﬁed,

which represent a new valid model state.

MODELSWARD 2016 - 4th International Conference on Model-Driven Engineering and Software Development

414

3 MATERIAL SEMANTICS

Material semantics generally covers enterprise-

related domains (Harel and Rumpe, 2004; Lindland

et al., 1994), which cannot be described with formal

syntax per se (Pfeiffer and Gehlert, 2005; Gehlert,

2007). This is of primal relevance in EM, as the entire

formalization of enterprise-related domains is practi-

cally not possible (cf. (Searle, 1984, p. 6), (Gehlert,

2007, p. 29) and (Pfeiffer and Gehlert, 2005, pp. 111-

112)). The reason for that lies in the lack of distinc-

tion between particular terms; especially regarding

synonyms, which cannot be differentiated conﬁdently

and cause a not Turing-computable problem (Pfeif-

fer and Gehlert, 2005). Besides, different EML users

usually have different understandings of EML meta

concepts (van der Linden, 2015).

We divide semantics on meta model layers into the

semantic domain and semantic mappings (Karagian-

nis and K

uhn, 2002)

. This understanding is shared

by numerous researches in the ﬁeld of ontological

analysis (Wand and Weber, 1993; Rosemann et al.,

2004).

3.1 Material Semantic Domain

The semantic domain covers all kinds of things the

language should be able to express (Harel and Rumpe,

2004, p. 68) and is conceptualized by domain con-

structs representing those things (Lindland et al.,

1994, p. 44). Consequently, syntactic constructs are

already pre-conceptualized and implicitly limit possi-

ble semantic references (Bjekovic et al., 2013; Wys-

susek, 2004). Basically, material semantics covers

what should be represented, i.e. those referred do-

main constructs that should be expressed (Esswein

and Weller, 2007, p. 2004). A further distinction into

static and behavioral material semantics as applied

in the ﬁeld of formal semantics (Hausmann et al.,

2005) seems to be superﬂuous, as the actual impact

of a syntactical construct becomes only relevant in

its individual application (cf. pragmatic semantics

(Bjekovic et al., 2014)). The underlying conceptu-

alization (Guizzardi, 2007) is only the carrier or inter-

mediate layer for that. In contrast, in formal seman-

tics the semantics itself represents the application, in-

dependent of any intermediate layer.

Domain constructs can represent physical things

as well as artiﬁcially constructed things of a domain

(Malt, 1990; Bloom, 1998). Thereby, it is extremely

We omit the consideration of semantics on the model

layer, which is referred as inherent semantics (Bork and Fill,

2014, p. 3403), since this paper explicitly aims to investi-

gate semantics within the meta model layer.

important to note that the perception, cognition and

interpretation of those things are subject-dependent

and determined by multiple factors (Bjekovic et al.,

2014; Esswein and Lehrmann, 2013; van der Linden,

2015).

We refer to this as structural ambiguity, as it ﬁ-

nally refers to the addressed domain constructs. Some

EM researches investigate dimensions of basal meta

concepts from the enterprise domain and demonstrate

their immanent variance (van der Linden and Hoppen-

brouwers, 2012; van der Linden and van Zee, 2014;

van der Linden, 2015). However, most EMLs implic-

itly base on positivistic assumptions (Bjekovic et al.,

2014, p. 438) indicating a precise and invariant refer-

ence to domain concepts, which is insufﬁcient in this

regard. Rather the positions of constructivism (mod-

els as constructions of reality-related conceptualiza-

tions) or at least critical rationalism (models as re-

constructions) seem to be more appropriate (Gehlert

et al., 2005; Esswein and Lehrmann, 2013) in order

to correspond to the immanent socio-pragmatic pur-

pose of EMLs (Bjekovic et al., 2014, p. 433).

3.2 Ambiguous Interpretation

Current representation approaches address the actual

description of material domains with the help of de-

notational approaches (K

uhn, 2004, p. 34); namely

pre-deﬁned generic ontologies (Wand and Weber,

1993) or EM-related ontologies (Gehlert et al., 2005;

offerer, 2007; Anaya et al., 2010; Fill, 2015). All

these approaches contain natural language statements

implying the lack of invariant interpretation (Thal-

heim, 2012). We refer to this as lexical ambiguity.

It covers same understandings with differing labeling

(e.g., synonyms) as well as different understandings

with same labeling (e.g., homonyms).

The stated ambiguities implicate subject-speciﬁc

understandings of material semantics, which is also

referred as personal semantics (van der Linden,

2015). We refer to this as actual semantics cov-

ering the actual understanding of an EML (perhaps

inﬂuenced by situational particularities (Henderson-

Sellers, 2005; Braun and Esswein, 2014). In contrast,

stipulated semantics constitutes as the actual seman-

tics of the language designers, which hence act as ini-

tial base for language understanding in the sense of an

original reference point for comparison (Rosemann

et al., 2004).

Reducing the mentioned difference between stip-

ulated and actual semantics of different stakehold-

ers causes the issue of semantic consensus-ﬁnding

(Harzallah et al., 2012, p. 501) and ontological com-

mitment (Guizzardi, 2007, p. 25) within EMLs. Frank

Towards Hybrid Semantics of Enterprise Modeling Languages

415

(2013) claims that the semantics of domain-speciﬁc

EMLs need to be invariant within a particular do-

main over a certain time period (Frank, 2013, pp. 13-

14). Obviously, such common understandings seem

to be realizable within a small modeler community.

It becomes more difﬁcult when addressing general-

purpose languages (e.g., UML) of purpose-speciﬁc

languages (e.g., BPMN), which are underspeciﬁed

and intentionally generic for a certain degree by deﬁ-

nition (cf. the Pool meta class in BPMN, for instance).

Consequently, respective syntactical constructs own

several semantic mappings. It becomes obvious that

respective consensus ﬁnding approaches - on top of

current representation formats - need to be elaborated

and investigated in further research.

3.3 Material Semantic Mapping

Semantic mapping is the mapping of syntactical con-

structs to domain constructs (Karagiannis and K

uhn,

2002). This mapping usually constitute as interpretive

step (Wand and Weber, 1993) due to the immanent

role of subject-dependent interpretation and missing.

In EM relevant languages, the mapping is mostly in-

formally speciﬁed (Bork and Fill, 2014, p. 3407). The

ﬁeld of ontological analysis of EMLs focuses on those

mappings in order to evaluate completeness of a lan-

guage in regard of a given ontology. For instance, the

BWW ontology provides a projection for the mapping

of syntactical concepts of EMLs by adapting the on-

tology of Bunge (Wand and Weber, 1993).

4 TOWARDS HYBRID

SEMANTICS

4.1 Pragmatics and Motivation

Pragmatics is understood as the actual application of

an EML in a speciﬁc context in order to solve or sup-

port a particular task (Lindland et al., 1994; Thalheim,

2012). Pragmatics in the ﬁeld of EM is rarely investi-

gated so far (Bjekovic et al., 2014). We hence propose

to distinguish pragmatics in accordance to the recip-

ient of the model that represents the major point of

utility creation, since enterprises constitute as socio-

technical information systems. Usually, the recipient

is a human actor, who is expected to perform some-

thing by using the model. This could be done for rea-

sons of pure documentation, which focuses the status

quo, or in regard of design and engineering, which

rather focuses desired states (e.g. designing a particu-

lar application system). If the recipient is a machine,

then this machine is expected to solve a particular

task that was technically speciﬁed before (e.g. model-

based analysis). Respective value is hence basically

created either by humans (variant, not determinable)

or machines (invariant, determinable). Potential hy-

brids between both types have to be decomposed un-

til a clear differentiation is possible. The creation of

value requires the understanding of models and re-

spective meta models. In case of expected human ac-

tions, material semantics are appropriate, while for-

mal semantics are relevant for machine actions. For-

mal semantics can be also relevant for human actors,

if interpretation invariance can be created based on

some kind of consensus ﬁnding.

There may be cases in EM, where both human ac-

tors and machines are recipient of a speciﬁc model.

For instance, in the ﬁeld of clinical process manage-

ment, clinical pathway models represent the process-

oriented, multi-perspective and integrated view on

treating patients and respectively required service

processes (Braun et al., 2014). While the treatment

process is naturally not formal, some service pro-

cesses can be formally deﬁned in order to execute au-

tomatable tasks. For instance, a particular resource

has to be automatically reordered by an ERP system

if the execution of a speciﬁc activity is completed. In

contrast, this activity is part of a treatment process

that is manually executed and the stated resource can

have different (material) semantics for human actors:

A physician perceives some resource as instrument

for treatment and a controlling expert refers to cer-

tain ﬁnancial aspects. Despite the different semantic

references of each actor (ERP system, physician, con-

trolling expert), the thing resource is mutual and pro-

vides a point for integration.

Therefore, it seems to be promising to discuss the

speciﬁcation of respectively required semantics in or-

der to enable an integrated deﬁnition of semantics (cf.

Section 1.2) and support the identiﬁcation of such se-

mantic integration points.

4.2 Framework

It becomes obvious that single semantic types are use-

ful for different purposes in the context of EM. These

types of semantics are therefore summarized in the

light of EM by focusing their semantics and ontolog-

ical positions concerning the real-world in Table 1.

They are further presented in detail below. Thereby,

also potential adaptations of formal semantics to orig-

inally non-formal issues from EM are brieﬂy consid-

ered.

Behavioral formal semantics can be seen as ap-

propriate means for the speciﬁcation of executable

MODELSWARD 2016 - 4th International Conference on Model-Driven Engineering and Software Development

416

Table 1: Types of Semantics in the Context of EMLs.

Behavioral Formal Seman. Static Formal Semantics Material Semantics

Principle Transitions between two

valid states by mapping to a

formal system and specifying

the transformation rules.

Deﬁnition of validation

rules for a model state

by mapping to a formal

system.

Mapping syntactical con-

structs to subject-dependent

domain constructs.

Purpose

for EMLs

Execution of parts of the

EML, i.e. some of its

meta classes become exe-

cutable. For instance, in-

variant machine operations

(Fill, 2015) or the semantics

of executable BPMN classes

(OMG, 2011, p. 435).

Restrictions on single meta

classes or constructs of

classes; similar to syntactic

semantics (Bjekovic et al.,

2014, p. 439). Alternatively

useful for on top model

analysis, i.e. mapping to a

formal target model.

Shaping a particular dis-

course and its basic concep-

tions (Proper et al., 2005);

pragmatic semantics in the

sense of relating syntax

to individual conceptualiza-

tions and the context of

a particular user (Bjekovic

et al., 2014).

Invariance Needs to be invariant and

mapped to truth-apt, exe-

cutable syntax.

Invariant for syntacti-

cal mapping, variant for

ontological mapping.

Variant to differing degrees

(cf. Section 3.2).

meta classes of an EML by specifying legal transfor-

mations between respective model states. It is tra-

ditionally useful for the speciﬁcation of any kind of

transformative and behavioral aspect in parts. This

might be the case for accompanying aspects of pro-

cess models, like resource-related aspects (Braun and

Esswein, 2014, p. 49). More precisely, the ﬂow

through a particular process model might cause hu-

man decisions and is hence not invariantly inter-

pretable. But the affected resource transformations

could be invariantly interpretable if particular re-

source allocations or transformation processes are

triggered and executed (see above). Although this

could be also done by an integration of a separate for-

mal language, it is more comprehensive to keep it in

one EML deﬁnition and one EML semantics deﬁni-

tion. Respective classes are therefore not solely de-

scriptive, but contain executable semantics.

Behavioral formal semantics require invariant se-

mantic mappings to a formal target model (syntax of

the explaining modeling language, cf. Section 2.3).

This indicates that a respective syntactical construct

from the meta model should become subject of an on-

tological discussion in order to ﬁnd respective con-

sensus. After reaching an invariant semantic mapping

to a particular domain construct, the domain construct

should be mapped invariantly to a syntactical con-

struct in order to specify respective transformations.

Brieﬂy speaking, the syntactical construct is mapped

to the syntactical construct of an other language (Fill,

2015) after an explicit agreement on the respective

reference domain is reached (Frank, 2013, pp. 13-

14). Finally, a particularly instantiated model rep-

resents not only real-world things descriptively, but

rather can manipulate parts of it (e.g., by coupling a

Enterprise Modeling Language

Hybrid Semantics

Material Domain

Material Semantics

Formal Syntax

Formal Semantics

Behavioral

Semantics

Static Semantics

Material Semantic Mapping

(variant interpretation)

Formal Syntax’

Formal Syntax’’

Formal Sem. Mapping

(invariant)

map

real-world interpretation

Figure 1: Outlining Hybrid Semantics in the Context of

EML Speciﬁcations.

particular machine control unit with respective model

elements).

Static formal semantics might be applicable for

the transformation of parts of the meta model into

other formal models aiming to deﬁne particular anal-

ysis on top of a model (external view) on the one side

(Natschl

ager, 2011a; Pietsch et al., 2014). On the

other side, it can be used for specifying valid model

states (internal view), which represent inherent rules

and restrictions of a particular domain.

In contrast to behavioral formal semantics, trans-

formation rules are not deﬁned. Due to the required

invariance between the syntactical constructs, respec-

tive consensus on the those relations is necessary,

Towards Hybrid Semantics of Enterprise Modeling Languages

417

i.e. consensus on the syntactical meaning of meta

model elements. This implies that a particular meta

model element should only be invariant in regard of

this relation, but can have several ontological map-

pings besides. Static formal semantics is hence view-

dependent.

Material semantics are naturally useful for the ex-

plication of individual conceptualizations about par-

ticular discourse objects (Guizzardi, 2007; Bjekovic

et al., 2014), indicating a particular agreement on re-

lated domain constructs (Lindland et al., 1994; Es-

swein and Lehrmann, 2013; van der Linden, 2015),

which aims to enable a common and efﬁcient com-

munication about a domain. Material semantics are

therefore primarily useful for supporting externaliza-

tion, documentation and communication between in-

dividuals.

Figure 1 outlines the concept of hybrid semantics

by splitting the EML semantics deﬁnition into parts

for material and formal semantics, which are inte-

grated by corresponding syntactical elements of the

meta model. The framework also depicts the ﬁnal

relation of the type of formal semantics to the real-

world.

5 CONCLUSIONS AND FURTHER

RESEARCH

This position paper outlines the prospective deﬁnition

of hybrid semantics for EMLs in order to explicitly

integrate formal and material semantics in a compre-

hensive manner. The paper should be understood as

starting point and contributes to the research com-

munity by structuring and framing different seman-

tic approaches and understandings, which are rele-

vant for EM. This could be seen as a ﬁrst step for

multi-disciplinary integration, tearing down still ex-

isting academic stonewalls between Information Sys-

tems Research (especially material semantics) and

Software Engineering (especially formal semantics)

in order to elaborate respective synergies for EML se-

mantics. This explicitly covers the aspect of using

and applying enterprise models and addresses ﬁnally

pragmatics, which is still a poorly investigated topic

that is largely omitted (Bjekovic et al., 2014).

Naturally, several topics need to be investigated in

further research. We explicitly propose the following

three topics.

Firstly, it is extremely important to specify the

considered integration of material and formal seman-

tics on the meta model layer. We therefore conduct

two case studies in the ﬁeld of manufacturing and

healthcare in order to derive more insights and tech-

nical consequences on that issue.

Secondly, it is also important to determine a spec-

iﬁcation format for hybrid semantics. It is there-

fore planned to conduct the DMM approach (Engels

et al., 2000) for formal aspects and ontology-based

approaches for material aspects. A structural en-

hancement of these ontologies should be examined

(van der Linden, 2015).

Thirdly, it is necessary to address the actual pro-

cess of consensus ﬁnding as stated in Section 3.2,

since a certain level of invariance is immanent for

any kind of execution and automation. This includes

the so far less discussed issue of explicitly specify-

ing semantics of generic EMLs for speciﬁc contexts,

domains or situations. This could lead to language

dialects keeping the syntax stable and changing the

semantic mappings, for instance (Braun and Esswein,

2014, pp. 47, 53).

ACKNOWLEDGEMENT

This research paper presents partial research results

from the research project SFB Transregio 96, which is

funded by the German Research Foundation (DFG).

REFERENCES

Anaya, V., Berio, G., Harzallah, M., Heymans, P., Mat-

ulevi

cius, R., Opdahl, A. L., Panetto, H., and Verde-

cho, M. J. (2010). The uniﬁed enterprise modelling

language—overview and further work. Computers in

Industry, 61(2):99–111.

Bjekovic, M., Proper, H. A., and Sottet, J.-S. (2013). En-

terprise modelling languages-just enough standardisa-

tion? In BMSD, pages 1–23.

Bjekovic, M., Proper, H. A., and Sottet, J.-S. (2014). Em-

bracing pragmatics. Conceptual Modeling, 8824:431–

444.

Bloom, P. (1998). Theories of artifact categorization. Cog-

nition, 66(1):87–93.

Bork, D. and Fill, H.-G. (2014). Formal aspects of enter-

prise modeling methods: A comparison framework.

In Proceedings of the 47th Hawaii International Con-

ference on System Sciences, pages 3400–3409. IEEE.

Braun, R. (2015). Towards the state of the art of extending

enterprise modeling languages. In Proceedings of the

3rd International Conference on Model-Driven Engi-

neering and Software Development, pages 394–402.

Braun, R. and Esswein, W. (2014). Classiﬁcation of

domain-speciﬁc bpmn extensions. In The Practice of

Enterprise Modeling, volume 147 of Lecture Notes of

Business Information Processing, pages 42–57.

MODELSWARD 2016 - 4th International Conference on Model-Driven Engineering and Software Development

418

Braun, R., Schlieter, H., Burwitz, M., and Esswein,

W. (2014). Bpmn4cp: Design and implementa-

tion of a bpmn extension for clinical pathways. In

IEEE International Conference on Bioinformatics and

Biomedicine (BIBM), pages 9–16. IEEE.

Burwitz, M., Schlieter, H., and Esswein, W. (2013). Mod-

eling clinical pathways-design and application of a

domain-speciﬁc modeling language. Wirtschaftsinfor-

matik Proceedings.

Cengarle, M. V., Gr

onninger, H., and Rumpe, B. (2014).

System model semantics of class diagrams. arXiv

preprint arXiv:1409.6635.

Delfmann, P., Herwig, S., and Lis, L. (2009). Uniﬁed

enterprise knowledge representation with conceptual

models-capturing corporate language in naming con-

ventions. ICIS 2009 Proceedings, page 45.

Engels, G., Hausmann, J. H., Heckel, R., and Sauer, S.

(2000). Dynamic meta modeling: A graphical ap-

proach to the operational semantics of behavioral di-

agrams in uml. In UML 2000 – The Uniﬁed Model-

ing Language, pages 323–337. Springer Berlin Hei-

delberg.

Engels, G., K

uster, J. M., Heckel, R., and Groenewegen,

L. (2001). A methodology for specifying and analyz-

ing consistency of object-oriented behavioral models.

In ACM SIGSOFT Software Engineering Notes, vol-

ume 26, pages 186–195. ACM.

Esswein, W. and Lehrmann, S. (2013). About the need for

semantically enriched reference models. In Proceed-

ings of the 19th Americas Conference on Information

Systems.

Esswein, W. and Weller, J. (2007). Method modiﬁcations

in a conﬁguration management environment. In Pro-

ceedings of the Fifteenth European Conference on In-

formation Systems, pages 2002–2013.

Fill, H.-G. (2015). Metamodeling as an interface between

syntax and semantics. In Gl

uck, B., Lachmayer, F.,

Schefbeck, G., and Schweighofer, E., editors, Elektro-

nische Schnittstellen in der Staatsorganisation, pages

43–50.

Frank, U. (2013). Domain-speciﬁc modeling languages: re-

quirements analysis and design guidelines. In Domain

Engineering, pages 133–157. Springer.

Frank, U. (2014). Enterprise modelling: The next steps.

Enterprise Modelling and Information Systems Archi-

tectures, 9(1):22–37.

Gehlert, A. (2007). Migration fachkonzeptueller Modelle.

Logos-Verlag.

Gehlert, A., Buckmann, U., and Esswein, W. (2005). On-

tology based method engineering. AMCIS 2005 Pro-

ceedings, page 436.

Gehlert, A. and Esswein, W. (2006). Toward more rigor

in ontological analyses. In Proceedings of the 14th

European Conference on Information Systems, pages

984–994.

Guizzardi, G. (2007). On ontology, ontologies, conceptu-

alizations, modeling languages, and (meta) models.

Frontiers in artiﬁcial intelligence and applications,

155:18.

Guizzardi, G., Ferreira Pires, L., and Van Sinderen, M. J.

(2002). On the role of domain ontologies in the design

of domain-speciﬁc visual modeling langages.

Hamann, L. and Gogolla, M. (2013). Endogenous meta-

modeling semantics for structural UML 2 concepts.

In Moreira, A., Sch

atz, B., Gray, J., Vallecillo, A., and

Clarke, P. J., editors, Model-Driven Engineering Lan-

guages and Systems - 16th International Conference,

MODELS 2013, Miami, FL, USA, September 29 - Oc-

tober 4, 2013. Proceedings, volume 8107 of Lecture

Notes in Computer Science, pages 488–504. Springer.

Harel, D. and Rumpe, B. (2004). Meaningful model-

ing: what’s the semantics of semantics? Computer,

37(10):64–72.

Harzallah, M., Berio, G., and Opdahl, A. L. (2012).

New perspectives in ontological analysis: Guidelines

and rules for incorporating modelling languages into

ueml. Information Systems, 37(5):484–507.

Hausmann, J. H. (2003). Metamodeling relations-relating

metamodels. Metamodelling for MDA, pages 147–

161.

Hausmann, J. H., Heckel, R., and Sauer, S. (2005). Dynamic

Meta Modeling. PhD thesis, University of Paderborn.

Heaven, W. and Finkelstein, A. (2004). Uml proﬁle to

support requirements engineering with kaos. IEE

Proceedings-Software, 151(1):10–27.

Henderson-Sellers, B. (2005). Uml–the good, the bad or the

ugly? perspectives from a panel of experts. Software

and Systems Modeling, 4(1):4–13.

offerer, P. (2007). Achieving business process model in-

teroperability using metamodels and ontologies. In

Proceedings of the 15th European Conference on In-

formation Systems, pages 1620–1631.

Holten, R. (2003). Integration von informationssystemen.

Wirtschaftsinformatik, 45(1):41–52.

Karagiannis, D. and K

uhn, H. (2002). Metamodelling plat-

forms. In EC-Web, volume 2455, page 182.

Kossak, F., Illibauer, C., Geist, V., Kubovy, J., Natschl

ager,

C., Ziebermayr, T., Kopetzky, T., Freudenthaler, B.,

and Schewe, K.-D. (2014). A Rigorous Semantics for

BPMN 2.0 Process Diagrams. Springer.

uhn, H. (2004). Methodenintegration im Business Engi-

neering. PhD thesis, Universit

at Wien.

Lam, V. S. (2012). A precise execution semantics for bpmn.

IAENG International Journal of Computer Science,

39(1):20–33.

Lankhorst, M. (2009). {Enterprise Architecture at Work:

Modelling, Communication and Analysis (The Enter-

prise Engineering Series)}.

Lankhorst, M. M., Proper, H. A., and Jonkers, H. (2009).

The architecture of the archimate language. In En-

terprise, Business-Process and Information Systems

Modeling, pages 367–380. Springer.

Lindland, O. I., Sindre, G., and Solvberg, A. (1994). Under-

standing quality in conceptual modeling. IEEE Soft-

ware, 11(2):42–49.

Maes, A. and Poels, G. (2007). Evaluating quality of con-

ceptual modelling scripts based on user perceptions.

Data & Knowledge Engineering, 63(3):701–724.

Towards Hybrid Semantics of Enterprise Modeling Languages

419

Malt, B. C. (1990). Features and beliefs in the mental repre-

sentation of categories. Journal of Memory and Lan-

guage, 29(3):289–315.

Maoz, S., Ringert, J. O., and Rumpe, B. (2010). A mani-

festo for semantic model differencing. In Models in

Software Engineering - Workshops and Symposia at

MODELS 2010, Reports and Revised Selected Papers,

pages 194–203.

Messer, B. (1999). Zur interpretation formaler

gesch

aftsprozeß-und workﬂow-modelle. In

Wirtschaftsinformatik und Wissenschaftstheorie,

pages 95–123. Springer.

Natschl

ager, C. (2011a). Deontic bpmn. In Database

and Expert Systems Applications, pages 264–278.

Springer.

Natschl

ager, C. (2011b). Towards a bpmn 2.0 ontology.

In Business Process Model and Notation, pages 1–15.

Springer.

OMG (2011). Business Process Model and Notation

(BPMN) - Version 2.0. Object Management Group

(OMG).

Opdahl, A. L., Berio, G., Harzallah, M., and Matulevicius,

R. (2012). An ontology for enterprise and information

systems modelling. Applied Ontology, 7(1):49–92.

Opdahl, A. L. and Henderson-Sellers, B. (2002). On-

tological evaluation of the uml using the bunge–

wand–weber model. Software and Systems modeling,

1(1):43–67.

Overhage, P. D. S., Birkmeier, D. Q., and Schlauderer, S.

(2012). Quality marks, metrics, and measurement pro-

cedures for business process models. Business & In-

formation Systems Engineering, 4(5):229–246.

Pfeiffer, D. and Gehlert, A. (2005). A framework for

comparing conceptual models. In Proceedings of the

Workshop on Enterprise Modelling and Information

Systems Architectures (EMISA 2005), pages 108–122.

Pietsch, P., M

uller, K., and Rumpe, B. (2014). Model

matching challenge: Benchmarks for ecore and bpmn

diagrams. arXiv preprint arXiv:1408.5693.

Proper, H. A., Verrijn-Stuart, A. A., and Hoppenbrouwers,

S. (2005). On utility-based selection of architecture-

modelling concepts. In Proceedings of the 2nd Asia-

Paciﬁc conference on Conceptual modelling-Volume

43, pages 25–34. Australian Computer Society, Inc.

Renger, M., Kolfschoten, G. L., and de Vreede, G.-J.

(2008). Challenges in collaborative modeling: A liter-

ature review. In Advances in Enterprise Engineering

I, pages 61–77. Springer.

Rosemann, M., Green, P., and Indulska, M. (2004). A

reference methodology for conducting ontological

analyses. In Conceptual Modeling, pages 110–121.

Springer.

Rospocher, M., Ghidini, C., Seraﬁni, L., Kump, B., Pam-

mer, V., Lindstaedt, S. N., Faatz, A., and Ley, T.

(2008). Collaborative enterprise integrated modelling.

In SWAP, volume 426. Citeseer.

Santos, P. S., Almeida, J. P. A., and Guizzardi, G. (2013).

An ontology-based analysis and semantics for organi-

zational structure modeling in the aris method. Infor-

mation Systems, 38(5):690–708.

Scheer, A.-W. and N

uttgens, M. (2000). ARIS architecture

and reference models for business process manage-

ment. Springer.

Schobbens, P.-Y., Heymans, P., and Trigaux, J.-C. (2006).

Feature diagrams: A survey and a formal seman-

tics. In Requirements Engineering, 14th IEEE inter-

national conference, pages 139–148. IEEE.

Searle, J. R. (1984). Minds, brains and science. Harvard

University Press.

Soltenborn, C. and Engels, G. (2009). Towards test-driven

semantics speciﬁcation. In Model Driven Engineering

Languages and Systems, 12th International Confer-

ence, MODELS 2009, Proceedings, pages 378–392.

Tarski, A. (1936). Der wahrheitsbegriff in den formal-

isierten sprachen.

Thalheim, B. (2012). Syntax, semantics and pragmatics of

conceptual modelling. In Natural Language Process-

ing and Information Systems, pages 1–10. Springer.

van der Linden, D. (2015). Personal semantics of meta-

concepts in conceptual modeling languages.

van der Linden, D. and Hoppenbrouwers, S. (2012). Chal-

lenges of identifying communities with shared seman-

tics in enterprise modeling. In The Practice of Enter-

prise Modeling, pages 160–171. Springer.

van der Linden, D. and van Zee, M. (2014). On the semantic

feature structure of modeling concepts: an empirical

study. In Business Informatics (CBI), 2014 IEEE 16th

Conference on, volume 2, pages 158–165. IEEE.

Vernadat, F. B. (2003). Enterprise modelling and integra-

tion. Springer.

Wand, Y. and Weber, R. (1993). On the ontological expres-

siveness of information systems analysis and design

grammars. Information Systems Journal, 3(4):217–

237.

Wand, Y. and Weber, R. (2002). Research commen-

tary: information systems and conceptual modeling—

a research agenda. Information Systems Research,

13(4):363–376.

Wyssusek, B. (2004). Ontology and ontologies in informa-

tion systems analysis and design: A critique. AMCIS

2004 Proceedings, page 535.

Wyssusek, B. (2006). On ontological foundations of con-

ceptual modelling. Scandinavian Journal of Informa-

tion Systems, 18(1):article 9.

MODELSWARD 2016 - 4th International Conference on Model-Driven Engineering and Software Development

420