Towards an Ontology-driven Framework for Workﬂow Analysis

Hlomani Hlomani and Deborah Ann Stacey

University of Guelph, School of Computer Science, 50 Stone Road East, Guelph, Ontario, Canada

Keywords:

Ontologies, Workﬂow, Process Models, Workﬂow Modelling, Workﬂow Analysis, Ontology-based Reason-

ing, Knowledge-based Systems, Ontology-driven Compositional Systems.

Abstract:

Workﬂow management and the the whole ﬁeld of business process management has seen a lot of research

interest. This interest has evolved from the initial quest to automate manufacturing processes to the formaliza-

tion of process models. The reason for this interest can arguably be attributed to the fact that process models

form the core of workﬂow management systems. A plethora of modelling languages and notations have been

created through the years, albeit with dominance of proprietary languages that has been argued to be lack-

ing in terms of having formal semantics. The informal languages have seen more adoption at the expense

of those that are termed “academic languages” even though academic languages are believed to be more for-

mal. This paper considers the aspects of model transformation with the intension to bridge the gap between

modelling and analysis. The paper proposes a semantic approach (using ontologies) to both the mapping and

transformation of business process models written in one language (source) to another (target).

1 INTRODUCTION

The workﬂow concept has existed for many years.

While its origins are not from computer science, it

is now as much a computing concept as computa-

tional analysis is. In its initial conception, the concern

was the automation of manufacturing processes. We

now see a redeﬁnition of the workﬂow concept mostly

based on the premise that, scientists mostly in col-

laboration with computing expects have and continue

to develop computational models (i.e. simulations of

real world etc.) which can be strung together to form

well balanced systems. While workﬂow modelling

is at the core of workﬂow management and analy-

sis, there exist a plethora of modelling techniques.To-

date and for a long time, the modelling and speciﬁ-

cation of these work- ﬂows have always dependent

upon proprietary languages and systems which litera-

ture has argued the existence of some ﬂaws viz.: lack

of formal model basis, lack of support for patterns

(and are hence deﬁcient in terms of their expressive-

ness)(van der Aalst and Hofstede, 2002), rigidity of

the languages both at design and execution (Almeida

et al., 2004) etc.

Two camps of process modelling languages exists:

those that originated from industry and academic lan-

guages. Industry languages are believed to be widely

adopted and used in the industry while academic lan-

guages are said to be less appropriate for being used

in concrete industrial application domains. The indus-

try languages have the limitation of not having formal

bases (formal semantics), something which academic

languages have. It is then apparent that choice be-

tween which language to use has been a matter of

trade-off. For example, a choice between the easy

and intuitive Uniﬁed Modelling Language’s Activity

Diagram (UML AD) that has no formal basis versus

Petri Nets which have a theoretical bases but are much

more involving.

In light of the absence of that “one” leading nota-

tion or language, model transformation would play an

important role in the world of workﬂow modelling.

Model transformation would be concerned with the

translation of a workﬂow model described in a source

language to a semantically equivalent model in a tar-

get model. It is against this background that we pro-

pose a framework for workﬂow modelling and analy-

sis that utilizes an ontology-based knowledge base, a

semantic transformation function, and universal anal-

ysis engine.

2 BACKGROUND

2.1 Ontologies

A search for a precise and concise deﬁnition of “on-

405

Hlomani H. and Stacey D..

Towards an Ontology-driven Framework for Workﬂow Analysis.

DOI: 10.5220/0004165404050410

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2012), pages 405-410

ISBN: 978-989-8565-30-3

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

tology” especially as it relates to computing can be

a challenging task. This is perhaps because the con-

cept never really had its origin in computer science.

The concept of ontology has its origins in the ﬁeld

of philosophy where it refers to a branch of meta-

physics that endeavours to offer a systematic account

of existence (Gruber, 1993). In the computer science

domain, however, as deﬁned by prominent players in

the domain such as the Object Management Group

(OMG), World Wide Web Consortium (W3C) and pi-

oneers in the domain ( such as Tom Gruber ), an on-

tology is “a formal conceptualization of a domain of

interest” (Group, 2009; Consortium, 2009; Gruber,

1993). A conceptualization is in fact an abstraction

of that which we wish to represent. In this abstrac-

tion, a declarative formalism is used to specify the

objects in the domain, their desirable properties and

the relationships among them. This forms a shared

vocabulary with which knowledge-based systems can

be built.

2.2 Ontology-driven Compositional

Systems

Motivations for compositional systems abound, even

so for those that are ontology-driven. These composi-

tional systems are of varying types (ranging from web

services-based, agent-based, resource-based, grid-

based etc.). Some of these came about because of

the advent of web services that opened up a world of

opportunities for developers, service providers, ser-

vice consumers, and organizations. Web services are

networked capabilities that openly exposes interfaces

for other machines to discover, rank, bind to and in-

voke (Blake and Nowlan, 2008). Their characteristics

such as interoperability, network and platform inde-

pendence has sparked a lot of research interest with

researchers investigating the many ways in which a

varying number of web services can be put together

to form a well-balanced system (Charﬁ and Mezini,

2007). Researchers such as (Cardoso and Sheth,

2005; Kim et al., 2010) have focussed on the web

service composition approach that exploit ontologies

to provide descriptions of web services thereby aiding

in the discovery, selection, composition and execution

of web services. These type of systems are made pos-

sible by the deﬁnition of standards such as the Web

Ontology Language for Service (OWL-S) from W3C

(wide Web Consortium, 2004).

Another distributed platform that research has fo-

cused on (in the context of compositional systems) is

that of agent based platforms. A notable work in this

regard is that by a group of researchers at the Stan-

ford university who implemented a prototype system

called BIOSTORM (Nyulas et al., 2008). In this re-

search, the primary goal was the integration of various

heterogeneous data sources, the deployment of a var-

ied number of detection algorithms through the col-

laboration of multiple agents. Like the BIOSTORM

project, Wang et al. (Wang et al., 2005) proposed

an agent-based framework that utilizes ontologies as

their knowledge-base. While differing in their usage

and interpretation of the ontologies ( (Nyulas et al.,

2008) using ontologies to describe the data sources as

well as detection methods and (Wang et al., 2005) fo-

cusing on description of the workﬂow so as to make

its execution dynamic) the overall platform remain

relatively similar ( i.e. multi-agent platform).

Based on the premise that scientist mostly in col-

laboration with computing experts have created com-

putational models which may or may not be cast as

web services, or agents (Hlomani and Stacey, 2009;

Gillespie et al., 2011) focused on a ”plug and play”

bottom-up system composition approach. This ap-

proach leverages the user’s expert knowledge (cap-

tured in ontologies) and assist the user in the compo-

sition of a system through the connection of composi-

tion units described in the ontologies that underlie the

system.

In all these research, the ultimate result was some

sort of workﬂow. The limitation of most of these im-

plementations was that less focus was given to the

analysis of the workﬂow. If any analysis was per-

formed, it was limited to syntactic checks which one

may be inclined to say are trivial to detect.

2.3 Workﬂow Modelling

Business processes are the core of workﬂow man-

agement hence, the paramount need for proper

modelling, analysis and veriﬁcation (Salimifard and

Wright, 2001). Process models are an abstraction of

the business requirements in terms of executable tasks

that put together in the end produce an intended out-

put. It is worth mentioning that there exists many

different ways to model business processes (from the

simple but intuitive graph theory (mostly directed

graphs), to the more complex but somehow mechan-

ical Petri Nets) (Atsa et al., 2011). The question of

which technique to follow seems a difﬁcult one to an-

swer since each technique comes with its justiﬁcation

and possibly disclaimers as to what the limitations

are. Having said that, (Cardoso et al., 2006) argues

that the decision to use a certain modelling technique

could be determined by the complexity of the prob-

lem. Other reasons for using a certain technique could

be motivated by the expressiveness, and formal ba-

sis of the model (e.g. W-Nets and YAWL (van der

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

406

Aalst, 2003)), coverage of both local and global de-

pendencies (e.g. deterministic graphs and path con-

straints (Fan and Weinstein, 1999)) or the simple and

intuitive nature of control ﬂow graphs and the Uniﬁed

Modelling Language (UML) etc.

While these process modelling techniques may

differ in such things as their expressiveness, formal

basis etc., a relative relation can be drawn between

them (including the fact that most are graph-based or

follow some sort of a graph formalism and that some

are derivatives or extensions of the other).

2.4 Workﬂow Languages

In the studies leading to the proposal of YAWL

(van der Aalst and Hofstede, 2002) and later on

(van der Aalst, 2003) offer a comprehensive analy-

sis of workﬂow languages. These motivating factors

for YAWL are also echoed by (Weske et al., 2006).

In both cases they make two signiﬁcant observations

viz.: lack of formal bases (or at least no practical

evidence to that fact), and lack of support for work-

ﬂow patterns (or some patterns). In these analyses,

what is clear is that, despite efforts to standardize

by such organizations as the workﬂow management

coalition, consensus with regards to workﬂow lan-

guages is far from agreed. This is perhaps as a result

of the many ways in which business processes can be

described. Their work is biased towards Petri Nets

particularly High-Level Petri Nets (i.e. Petri Nets ex-

tended with colour, time, and hierarchy) which further

forms the foundation for YAWL. Workﬂow languages

were evaluated to determine their support for work-

ﬂow patterns and were found to be wanting in terms

of their expressive power.

Business Process Execution Language For Web

Services (BPEL4WS) offers another interesting case

for workﬂow languages. It is popular in the web ser-

vice composition domain (if not the de facto standard)

(Weske et al., 2006). BPEL4WS carries the blocked-

structure and graph-based modelling from its prede-

cessors ( XLang and Web Services Flow Language

(WSFL), respectively). While it is said to be lacking

a graphical notation, some constructs of the BPMN

discussed earlier can be directly mapped to BPEL and

hence can be executed in a BPEL execution engine.

2.5 Workﬂow Ontology

Some work has been done towards the ontological

representation of the workﬂow concept. One such

work is the IntelLEO Workﬂow Ontology (Jovanovic

et al., 2011). While the motivations and structure of

this ontology has a bias towards the description of the

learning ﬂows, it has a sense of “traditional” work-

ﬂows. This is because the ordering (either sequential

or parallel) of activities to achieve some goal can be

achieved. The ontology, however, lacks the expres-

sive ability to capture even some of the most basic

concepts of process models (e.g. the notions of rout-

ing beyond just “sequence” and “parallel”).

Jenz (Jenz, 2003) explores an interesting aspect of

software design: the generation of software artifacts

from an ontological knowledge base (KB). Much like

the approaches discussed in section 2.2, this approach

is dependent on the deﬁnition of a knowledge base (

ontology + instances = KB). Contrary to the previ-

ously discussed ontologies, the aim of the business

process ontology (Jenz, 2003) is to be generic in its

description of business processes. This is achieved

through a static representation of a business process

by focusing on activities or tasks as the “building

blocks”. The representation of the business process in

an ontology then achieve two main goals viz.: provide

a vendor-neutral and platform-independent descrip-

tion of the business process, and provide both human-

understandable and machine-readable description of

the business process. The ontology is general to some

extent ( it covers a larger number of the concepts

needed to describe business processes). However,

while synonyms and homonyms can be included in

the ontology to cater for terminology gaps, there are

cases where a one-to-one mapping cannot be achieved

particularly where business processes are concerned.

To give an example, BPMN, UML activity diagrams

and Petri Nets are all business process modelling tech-

niques. While we could say thet a transition (Petri Net

concept) is synonymous to say an action state (UML

concept) or a vertex (Graph concept), it has been ar-

gued that this may work for many of the concepts but

not for some constructs (Lohmann et al., 2008). The

separation of concerns principle would therefore, be

ideal. This would follow a pattern where top level

concepts are detached from a speciﬁc modelling tech-

nique and further specialized going downward. This

is opposed to the idea followed in this ontology where

there exists a bias towards the use of UML concepts.

3 THE PROPOSED

FRAMEWORK

In section 2.3 we mentioned that business process/

workﬂow modelling is a core element of the business

process management and workﬂow analysis task. We

also deduced from literature that there exist umpteen

modelling techniques. Sufﬁce it to say that each mod-

elling technique has its role to play in the business

TowardsanOntology-drivenFrameworkforWorkflowAnalysis

407

process modelling process (depending on the mod-

elling context, the problem in question in terms of

size and expressiveness requirement etc.). Having

said this, it also sufﬁces to say that choice of which

model to use depends entirely on the user, rather, on

the modeller.

Our proposal endeavours to ease the burden of

modelling technique choice by providing a means to

traverse the business process modelling space. In this

framework we focus on the transformation from one

model to the other with minimal loss of information.

This framework is depicted in ﬁgure 1. We identify

components of the framework to include: Ontology-

Driven Compositional System/ Workﬂow Manage-

ment System/Business Process Modelling System,

Knowledge base, Semantic Transformation, and Uni-

versal Analysis Engine (this is not discussed because

it will not be a part of the initial instance of the frame-

work). These are discussed in subsections that follow.

Figure 1: A universal framework for workﬂow management

and analysis. The framework relies on an ontology-based

knowledge base.

3.1 The Knowledge Base

The goal is to have a universal (at least in terms of

coverage) knowledge base for workﬂow representa-

tion/ modelling models. Following the notion that

process models are mostly similar with subtle differ-

ence ( a one-to-one mapping for some constructs is

not possible), the knowledge base would then have

a generic top-level description (describing the com-

mon constructs) and model speciﬁc constructs further

to the bottom (describing the constructs that are spe-

ciﬁc to modelling languages). The most general con-

stituents of a process model are activity, and relation.

In terms of relation, the modelling of split and join

elements is important because these facilitate the cre-

ation of workﬂow patterns (e.g. parallel routing etc.).

Figure 2 depicts the structure of such an ontology

in which clouds represents imported ontologies (these

would in most cases be addressing specialization con-

structs), rectangles represent the classes, arrows rep-

resents relationships, while the lines represents sub-

sumption (sub-classes). The most basic deﬁnition of

a workﬂow is a graph (ordered pair) G = (A, E) con-

sisting of a set A of Activities together with a set E

of Relations, the ontology is thus, representative of

this structure (at the top level). Having deﬁned these

general notions (workﬂow/process, activity, and con-

trol ﬂow), speciﬁc modelling techniques could then

be mapped to these top-level elements as exempliﬁed

in ﬁgure 2 by sub-classing the workﬂow class to de-

ﬁne a Petri Net sub-class and drawing a relation be-

tween the Petri Nets constructs and the workﬂow con-

structs (e.g. between the Activity class and Transition

class since these classes are semantically equivalent).

Figure 2: Graphical view of the workﬂow ontology. Like

the framework, the ontology has a generic yet extensible

structure to allow for further specialization through map-

pings to other ontologies and concepts whenever needed.

3.2 The Semantic Transformation

The semantic transformation of process models lever-

ages knowledge embodied in the knowledge base. In

the knowledge base, an explicit mapping is created

between each of the process models to the generic

workﬂow and thus an implicit mapping also exists

between the process models themselves. Lose ends

would probably exist as mentioned earlier that some

constructs may not be directly mapped.

Since this is envisioned to be a universal frame-

work, our postulation is that mechanisms will be

reused from other existing implementations. For ex-

ample, for a serialization format, standards such as

the Workﬂow Process Deﬁnition Language (WPDL)

(Junginger, 2000) used for the import and export of

workﬂows between workﬂow management systems

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

408

can be utilized or used as a basis. The actual map-

ping and transformation between process models (es-

pecially in cases where semantic mismatches exists)

can consider mappings from existing work such as

the WPDL standard mentioned earlier, mappings and

transformations by (Lohmann et al., 2008) as well as

formal and general deﬁnitions of the process mod-

elling techniques as provided in literature would serve

as input in the creations of the transformation mecha-

nism.

4 DISCUSSION

The most obvious beneﬁt of model transformation is

that business process modelling can be performed at

any level, and in any process modelling technique

(perhaps because of the technique’s simplicity or in-

tuitiveness as is the case with UML). This is because

you can model with a language of choice and then

convert to a more formal tool (e.g. petri nets) for

analysis. The potential beneﬁts of this framework em-

anates from a knowledge base (especially one that is

generic) can be reused and shared between projects.

Ontologies that describes a construct of interest for

example can always be imported. Such has been with

the case with prominent ontologies like the dublin

core and the friend-of-a-friend(FOAF). The frame-

work is also attractive because of the structure of the

ontology which promotes ﬂexibility. Whenever the

need to incorporate a new modelling technique arise

an ontology for that particular model can be deﬁned

and mappings then can be create between the con-

structs of that ontology and those of the generic on-

tology.

5 CONCLUSIONS

In this paper we have proposed a universal frame-

work for workﬂow analysis. We have discussed the

notions relating to the knowledge base for the frame-

work, its transform function and the universal analysis

engine. The framework was viewd to have attractive

traits such as the ﬂexible and extensible structure of

the knowledge base, opening of possibilities to model

in virtually any language of choice and the universal

coverage of the analysis engine. While this work is

preliminary, it would close gaps in the business pro-

cess modelling by offering a universal playground for

manipulation of workﬂows and business processes.

REFERENCES

Almeida, T., Vieira, S. C., and Casanova, M. A. (2004).

Flexible workﬂow execution through an ontology-

based approach. In Workshop on Ontologies as Soft-

ware Engineering Artifacts (OOPSLA).

Atsa, R., Marcel, E., Ndjodo, F., and Aloo, G. A. (2011).

A formal framework for business process model-

ing. International Journal of Computer Applications,

12(6):27–32.

Blake, M. and Nowlan, M. (2008). Taming web services

from the wild. IEEE Internet computing, 12:62–69.

Cardoso, J., Mendling, J., Neumann, G., and Reijers, H.

(2006). A discourse on complexity of process mod-

els. In Eder, J. and Dustdar, S., editors, Business Pro-

cess Management Workshops, volume 4103 of Lecture

Notes in Computer Science, pages 117–128. Springer

Berlin / Heidelberg.

Cardoso, J. and Sheth, A. (2005). Introduction to Semantic

Web Services and Web Process Composition, volume

3387 of Lecture Notes in Computer Science. Springer

Berlin Heidelberg, Berlin, Heidelberg.

Charﬁ, A. and Mezini, M. (2007). AO4BPEL: An Aspect-

oriented Extension to BPEL. World wide web,

10:309–344.

Consortium, W. W. W. (2009). OWL 2 Web Ontology

Language, Document Overview: W3C Recommenda-

tion 27 October 2009. W3C. http://www.w3.org/TR/

2009/REC-owl2-overview-20091027/.

Fan, W. and Weinstein, S. (1999). Specifying and reason-

ing about workﬂows with path constraints. In Hui, L.

and Lee, D.-L., editors, Internet Applications, volume

1749 of Lecture Notes in Computer Science, pages

25–54. Springer Berlin / Heidelberg.

Gillespie, M. G., Stacey, D. A., and Crawford, S. S.

(2011). Designing Ontology-Driven System Composi-

tion Knowledge and Processes to Satisfy User Expec-

tations (in publication). Lecture Notes in Computer

Science. Springer-Verlag.

Group, O. M. (2009). Ontology Deﬁnition Metamodel -

OMG Document Number: formal/2009-05-01. OMG.

Gruber, T. R. (1993). Toward principles for the design

of ontologies used for knowledge sharing. In Inter-

national Journal of Human-Computer Studies, pages

907–928. Kluwer Academic Publishers.

Hlomani, H. and Stacey, D. A. (2009). An ontology driven

approach to software systems composition. In Inter-

national Conference of Knowledge Engineering and

Ontology Development, pages 254–260. INSTICC.

Jenz, D. E. (2003). Simplifyingthe software development

value chain through ontology-driven software artifact

generation.

Jovanovic, J., Siadaty, M., Lages, B., and Spors, K. (2011).

IntelLEO workﬂow ontology.

Junginger, S. (2000). The workﬂow management coali-

tion standard wpdl: First steps towards formalization.

In 7

European Concurrent Engineering Conference

(ECEC2000), pages 1–6.

Kim, J., Gil, Y., and Spraragen, M. (2010). Principles for

interactive acquisition and validation of workﬂows.

TowardsanOntology-drivenFrameworkforWorkflowAnalysis

409

Journal of Experimental and Theoretical Artiﬁcial In-

telligence, 22(2):103–134.

Lohmann, N., Verbeek, E., and Dijkman, R. (2008). Petri

net transformations for business processes - a survey.

Transactions on Petri Nets and Other Models of Con-

currency.

Nyulas, C., OConnor, M., Tu, S., Buckeridge, D., Okhma-

tovskaia, A., and Musen, M. (2008). An ontology-

driven framework for deploying jade agent systems.

Salimifard, K. and Wright, M. (2001). Petri net-based mod-

elling of workﬂow systems: An overview. European

Journal of Operational Research, 134(3):664 – 676.

van der Aalst, W. (2003). Pi calculus versus petri nets: Let

us eat ”humble pie” rather than further inﬂate the ”pi

hype”.

van der Aalst, W. and Hofstede, A. H. M. T. (2002). Yawl:

Yet another workﬂow language. Information Systems,

30:245–275.

Wang, S., Shen, W., and Hao, Q. (2005). Agent based work-

ﬂow ontology for dynamic business process compo-

sition. International Conference on Computer Sup-

ported Cooperative Work in Design, 1.

Weske, M., Vossen, G., and Puhlmann, F. (2006). Work-

ﬂow and service composition languages. In Bernus,

P., Mertins, K., and Schmidt, G., editors, Handbook

on Architectures of Information Systems, International

Handbooks on Information Systems, pages 369–390.

Springer Berlin Heidelberg.

wide Web Consortium, W. (2004). OWL-S: Semantic

Mark-up for Web Services. ”http:// www.w3.org/

Submission/ OWL-S/”.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

410