Syntax highlighting in business process models.

Full text

Page 1

AUTHOR QUERY FORM
Journal: DECSUPPlease e-mail or fax your responses and any corrections to:
E-mail: corrections.esil@elsevier.spitech.com
Fax: +1 619 699 6721
Article Number: 11841
Dear Author,
Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted by flags in
the proof. Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-
screen annotation in the PDF file) or compile them in a separate list.
For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions.
Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted by flags in
the proof. Click on the ‘Q’ link to go to the location in the proof.
Location
in article
Query / Remark: click on the Q link to go
Please insert your reply or correction at the corresponding line in the proof
Q1Please check the telephone/fax number of the corresponding author, and correct if necessary.
Thank you for your assistance.
Our reference: DECSUP 11841P-authorquery-v8
Page 1 of 1

Page 2

1
Syntax highlighting in business process models
2
H.A. Reijersa,⁎, Thomas Freytagb, Jan Mendlingc, Andreas Ecklederb
3
4
5
aEindhoven University of Technology, The Netherlands
bBaden-Württemberg Cooperative State University Karlsruhe, Germany
cHumboldt-Universität zu Berlin, Germany
6
7
8
9
10
11
12
1314
15
16
17
18
19
20
21
a b s t r a c ta r t i c l ei n f o
Article history:
Received 2 March 2010
Received in revised form 12 November 2010
Accepted 24 December 2010
Available online xxxx
Keywords:
Business process models
Workflow nets
Understandability
Process modeling tool
Coloring
22
23
24
25
26
27
28
29
30
31
Sense-making of process models is an important task in various phases of business process management
initiatives. Despite this, there is currently hardly any support in business process modeling tools to adequately
support model comprehension. In this paper we adapt the concept of syntax highlighting to workflow nets, a
modeling technique that is frequently used for business process modeling. Our contribution is three-fold.
First, we establish a theoretical argument to what extent highlighting could improve comprehension. Second,
we formalize a concept for syntax highlighting in workflow nets and present a prototypical implementation
with the WoPeD modeling tool. Third, we report on the results of an experiment that tests the hypothetical
benefits of highlighting for comprehension. Our work can easily be transferred to other process modeling
tools and other process modeling techniques.
© 2011 Elsevier B.V. All rights reserved.
32 33
34
35
36
1. Introduction
37
Capturing business processes in the form of graphical models has
become a popular way to support the communication between
business professionals and to guide the development and implemen-
tation of IT systems [11,30]. An abundance of academic literature is
devoted to the formal aspects of process modeling (see e.g.
[34,46,61]), while much of the efforts in industry are geared towards
standardizing the involved notations. A good example of the latter is
the adoption of the Business Process Modeling Notation 2.0 as a
formal OMG standard at the end of 2009.
What have received comparatively little attention are the factors
that make the usage of process models effective. Because the primary
purpose of process models is to facilitate human communication and
problem solving – as is the case for most visual diagrams [27] – a key
issue is how to improve the understanding of such models. In other
words: How can process models be created such that they can be
understood more quickly and accurately by human model readers?
This question is of significant relevance. Many companies build their
process management initiatives on large-scale process model repos-
itories that often contain several thousands of process models [55,56].
Casual modelers and usual staff members create and read these
models in support of their daily operations. Currently, research shows
that there are serious issues with the creation and comprehension of
these models [40]. Since the ease with which process models can be
understood is shown to be a positive influence on the success of
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
projects that use these to improve business processes and develop
information systems [33], answers to the question of accurate
comprehension can be expected to directly benefit such projects.
This paper's focus is on the highlighting of syntactical elements in
process models to improve their understandability. More specifically,
we propose the use of color to highlight process model elements that
relate to one another in a way that is comparable to how pairs of
opening and closing brackets in a natural sentence do. The technique
of syntax highlighting, i.e. the coloring or emphasizing of source code
in meaningful ways, has become an established feature in program-
ming editors to support programmers in making sense of code.
Despite the similarities that have been noted between process models
and software code [26,65], syntax highlighting of process models has
not been introduced yet. This is all the more surprising given the wide
availability of process modeling tools. At this stage, the use of color in
process models is used mainly to distinguish between different types
of model elements, for example to categorize events (purple) and
functions (green) in Event-driven Process Chains [57]. However, color
is not used systematically to aid sense-making of specific process
models. Against this background, the contribution of this article is
three-fold. First, we provide a detailed discussion of potential benefits
of syntax highlighting for process model comprehension from a
theoretical perspective. Second, we formalize a highlighting concept
for workflow nets and demonstrate its applicability with an
implementation within the open source modeling tool WoPeD.
Third, and based on that implementation, we report the results of
an experiment that challenges the benefits of the highlighting
approach.
The structure of the paper is now as follows. In Section 2, we
discuss the effects of syntax highlighting using insights from cognitive
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
Decision Support Systems xxx (2011) xxx–xxx
⁎
Q1
Corresponding author. Tel.: +31 402473629; fax: +31 402432612.
E-mail address: h.a.reijers@tue.nl (H.A. Reijers).
DECSUP-11841; No of Pages 11
0167-9236/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.dss.2010.12.013
Contents lists available at ScienceDirect
Decision Support Systems
journal homepage: www.elsevier.com/locate/dss
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 3

91
92
research. Section 3 defines a formal approach to highlighting for
workflow nets and presents a corresponding implementation.
Section 4 describes an empirical test that has been conducted to
determine the effectiveness of the proposed highlighting approach.
We discuss implications of our work in Section 5 and conclude our
paper with Section 6.
93
94
95
96
97
2. Syntax highlighting in process models
98
99
In this section we introduce the theoretical background of our
research on syntax highlighting for process models. In Section 2.1 we
discuss the concept of secondary notation and its importance for
process model understanding. We use an example of a real-world
process model to illustrate the potential benefits of highlighting as a
mechanism of secondary notation. In Section 2.2 we investigate how
different user groups might benefit from highlighting.
100
101
102
103
104
105
2.1. Color and understanding
106
Traditionally, conceptual models including business process
models are created as an interplay between an expert in the
considered domain (domain expert) and an expert in modeling
techniques (system analyst) [20]. Typically, a domain expert can be
characterized as someone with (1) superior, detailed knowledge of
the object under consideration but often (2) minor powers of
abstraction beyond that knowledge. The strengths of the system
analyst are exactly mirrored. Recently, business process modeling
projectshave growntocompany-wide initiativesin whichnon-expert
modelers (or novice modelers) are increasingly active. Such projects
can easily cover the definition and maintenance of several thousands
of models. The trend towards an increasing involvement of novices in
process modeling projects causes various quality issues [56]. Recent
research reveals considerable weaknesses of process models from
practice in terms of understanding and error probability. Many real-
world process model collections have error rates of up to 30% [40].
Such quality issues have been partially attributed to the sheer
complexity of certain process models [39,41,42,45]. Therefore, it is
an important question how readers can be better supported in
understanding a process model in an accurate way.
Most process models and corresponding languages are rather
puristic from a visual point of view. Hardly any highlighting is used
except for Event-driven Process Chains in which sometimes events
(purple) and functions (green) are distinguished by color. The
research by Bertin on the semiology of graphics identifies eight
distinct visual variables that can be used to encode graphical
information [4]. Color is considered to be one of the most effective
of these variables. The human visual system is able to recognize color
quickly [38]. Distinctions between different colors can be detected
accuratelyandthreetimesfaster thanbetweenshapes[37]. Therefore,
Moody criticizes that color is hardly used by modeling notations to
distinguish notation elements [47]. These insights clearly point to the
attractiveness of using color in a systematic way to improve the
understanding of process models.
Furthermore, the Cognitive Dimensions Framework (CDF) by
Green and Petre provides the background to postulate a way to
sensibly apply coloring [23,24]. The CDF has been highly influential in
language usability studies and numerous publications have been
devoted to its further development, see its discussion in [5]. It
providesa setof characteristics toevaluatea widevarietyof notations,
e.g. spreadsheets, style sheets, diagrams, etc. Specifically, process
modeling languages have been analyzed to be “abstraction-hating”,
because they do not provide any mechanism to group activities. This
leads to several problems in terms of cognitive processing as it is not
directly visible where a certain sub-component of a process model
starts and where it ends. As a result, the inside of a component cannot
be ignored when considering behavioral aspects around it and it is
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
hard to neglect a sub-component's environment when its internal
aspects are investigated. Additionally, in terms of the CDF the start
and end nodes of sub-components are examples of long-range
“hidden dependencies”. In a process model, it is very common that
a particular type of start node calls for a similar type of end node.
However, the larger and more complex a model gets, the more
difficult it is to determine which start and end nodes belong to one
another. This is problematic since identifying matching entry and exit
nodes of a component helps the model reader in terms of information
hiding [50,54].
Motivated by the discussed shortcomings of process models as a
visual notation according to the CDF and the potential power of color
as a visual variable mentioned earlier, it seems attractive to focus on
the highlighting of entry and exit nodes of sub-components. Such
nodes typically reflect a particular routing semantics. In a workflow
net context, they are called operator transitions. We will use this term
already here; its precise definition will be part of the section on our
formal highlighting approach.
How highlighting may be used to facilitate information hiding is
illustrated in Fig. 1. It shows two versions of a real-world process
model that has been made available to us through a cooperation with
the Dutch branch of Sogeti,1a large ICT service provider. The process
captures how existing credit facilities are updated within one of
Sogeti's clients, a multinational bank. This is required at times, for
example because account holders demand updates or because the
commercial circumstances change. The process involves the execu-
tion of various checks by bank clerks, as well as updates they have to
carry out in IT systems. Sogeti professionals created a model for this
process to use it as a basis for discussion with bank employees on how
to improve the business process in question.
This process model is captured as a workflow net [60]. The model
captures 63 different transitions representing business activities (also
often referred to tasks), 75 places indicating milestones in the process,
and 157 arcs specifying the paths along these elements. Also notable
are the 30 operator transitions (XOR-splits, XOR-joins, AND-splits and
AND-joins), which capture how alternative and concurrent paths are
spawned off and joined again at different stages in the process. A
complex aspect of this model is that it is represented in a compact
form; the modelers wanted it to fit on a single sheet of A4 paper so
that it could be easily printed and reproduced. As a result, some
arrows are running bottom-up, which could easily be interpreted as
iterations in the model while this is not always so. For, example, the
arrow from B2 to F2 simply indicates sequential progress instead of a
step back.
While the formal structures of both versions of the process model
are the same, there is a notable difference in the way how matching
operator transitions are highlighted. Cognitive research into program
comprehension has coined the terms primary notation and secondary
notation to describe this phenomenon. The modeling notation as a
formal set of symbols is defined as primary notation. Primary notation
specifies the semantics of all graphical elements of a particular
notation, such as Petri nets. This primary notation of Petri nets is
defined using particular shapes for the different syntax elements. In
the model with highlighting, it can be seen that some sets of operator
transitions have received the same color, for example E1 and A2 that
are both highlighted in green. Transition operator E1 – an AND-split –
signifies that two concurrent paths are initiated after its execution,
one of them starting with transition H1, the other with transition I1.
By enriching the process model with information beyond the formal
notation (e.g. color, line strength, etc.), the reader may access the
information captured in themodelwitha differing degree ofease [51].
Visual cues, which are not part of a notation, are known as secondary
notation [52]. These visual cues have a twofold advantage. First, they
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
1http://www.sogeti.com/.
2
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 4

216
217
help to identify a decomposition of the process model into
components, which provides for information hiding when the overall
behavior of the process model is analyzed. Second, the usage of color
helps to interpret secondary notation quickly, as color can be
processed by the human visual system much faster than for instance
shape [37]. These facts should result in better comprehension
performance, which is typically measured in terms of accuracy and
efficiency [7,21]. Therefore, we formulate the following hypotheses in
relation to accuracy and efficiency of comprehension.
218
219
220
221
222
223
224
225
226
H1. The use of colors to highlight matching operator transitions will
have a significant positive impact on understanding accuracy.
227
H2. The use of colors to highlight matching operator transitions will
have a significant positive impact on understanding speed.
228
229
2.2. Highlighting, understanding, and expertise
230
Prior research has shown that users tend to have serious problems
with understanding how different operator transitions interplay and
which ones belong together. It comes as no surprise that model
readers with different characteristics face understanding problems to
a different extent. For instance, it has been observed that readers with
a solid background in Petri net concepts [42] and theoretical concepts
231
232
233
234
235
Fig. 1. Changing a credit facility process without (a) and with (b) highlighting.
3
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 5

236
237
of process models altogether [44] show a much better understanding
performance than others.Also, otherpersonal factorslike the duration
and intensity of process modeling experience have been identified as
factors of process model understanding in prior research [53].
Altogether, modeler expertise is a critical issue for process modeling
projects [3].
Archetypically, novices and experts in process modeling can be
distinguished. The notion of expertise has been related to different
aspects. As an adaptation of [10,16], we can state that it is established
by “the amount and complexity of knowledge gained through
extensive experience of activities” in process modeling [16] or by
acquiring “vast amounts of knowledge and the ability to perform
pattern-based retrieval” related to process models [10]. The difference
between novices and experts in process model comprehension can be
explained based on the Adaptive Control of Thought architecture
proposed by Anderson [2]. According to this architecture, the human
working memory interacts with declarative and production memory,
which both the latter have distinct features. While declarative
memory stores and provides access to facts that we explicitly know,
the production memory holds rules of interference, which can be used
in problem solving. The production memory of an expert contains a
much richer set of production rules than the one of a novice. In
relation to process models, an expert would likely know productions
that help to hide information, for instance, to ignore components
wheninterpreting the overall behaviorof a process model. Whileboth
novices and experts will presumably benefit from color highlighting,
it is likely that their differences in production memory will result to a
difference in the extent of this effect. Both an unhighlighted and a
highlighted process model are informationally equivalent, i.e. they
capture the same information on a process [35,58]. On the other hand,
they are not computationally equivalent because they differ in the
ease withwhichinformationcan be deducted fromthem.For a novice,
this computational benefit is great as a novice lacks suitable
production rules to inspect a process model. For an expert, the
computational improvement is smaller since she possesses produc-
tionrulestomanagethecomplexityof aprocessmodel.Therefore,itis
not a surprise that expert modelers have been observed to focus on
relevant graphical elements, recognize patterns and disregard
irrelevant information [51], while novices tend to lack reading and
search strategies, which result from modeling experience and
extensive learning. As a result, we expect highlighting of matching
nodes in a process model to be a significant aid for novices to read the
models and their control flow semantics. Experts, in turn, can much
easier identify patterns of matching operator transitions, a skill that is
also referred to as perceptual expertise [25], but they will nonetheless
benefit from the highlighting approach. Therefore, we formulate the
following hypotheses:
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
H3. The highlighting of matching operator transitions will have a
significant positive impact on understanding accuracy for novices.
285
286
H4. The highlighting of matching operator transitions will have a
significant positive impact on understanding speed for novices.
287
H5. The highlighting of matching operator transitions will have a
significant positive impact on understanding accuracy for experts.
288
289
H6. The highlighting of matching operator transitions will have a
significant positive impact on understanding speed for experts.
290
291
292
Before we can evaluate these hypotheses, we first have to
explicitly define how the syntax highlighting approach for process
models can work. This is the subject of the next section.
293
294
3. A formal approach to syntax highlighting
295
296
In this section we formalize a formal approach to highlighting
matching entry and exit nodes in nets that are inspired by so-called
workflow nets. Section 3.1 defines some Petri net concepts, which are
an important basis for the notion of workflow nets. Section 3.2
presents the highlighting approach as we have implemented it in
WoPeD.
297
298
299
300
301
3.1. Petri nets and workflow nets
302
303
A Petri net N=(P,T,F) is a directed, bipartite graph where P is a set of
nodes called places, T a set of nodes called transitions and Fp
(P×T)∪(T×P) a binary flow relation. For a node n∈P∪T of a Petri net,
we call •n = m∈P∪Tjðm;nÞ∈F
ðn;mÞ∈Fg the postset of n. Fig. 2 shows a Petri net with
P = p1;p2;p3;p4;p5
f
Here, for example, •t2= p2;p5
f
the postset of t3.
Asequenceofnodesπ = 〈n1; :::;nk〉ofaPetrinetwhere(ni,ni+1)∈F
for i∈[1..k−1] is called a path from n1to nk. A path π = 〈n1; :::;nk〉 is
called elementary path, if it does not contain the same node more than
once, i.e. for each two nodes niand njalways holds ni≠nj. In Fig. 2,
〈p1;t1;p3;t3;p4〉 is an example for an elementary path; 〈p1;t5;p2;t2;p1〉
isanexampleforanon-elementarypath.Sinceelementarypathscontain
eachvisitednodeonlyonce,theycanberepresentedasplainsetsinstead
of sequences. For technical reasons, we define an alphabet operator α
mapping an elementary path sequence to a plain set of nodes:
α 〈n1; :::;nk〉
ð
In the area of business process modeling, Petri nets have been used
as the basis for so-called workflow nets. By now, these have become
widely-used to capture business processes in a graphical form.
Workflow nets were originally introduced in [61] and have been
used in many applications and publications ever since. Inspired by the
more recent workflow net variant that is being presented in [60], a
workflow W net can be formally defined as a tuple (P,T,F,TAS,TXS,TAJ,
TXJ) with the following properties:
304
305
306
307
308
309
fg the preset of n and n• = m∈P∪Tj
f
g,
T = t1;t2;t3;t4
f
g is the preset of t2and t3• = p4
g,
F = ðp1;t1Þ;ðt1;p3Þ; :::
fg.
fg is
310
311
312
313
314
315
316
317
318
319
Þ = n1; :::;nk
fg.
320
321
322
323
324
325
326
327
328
329
• (P,T,F) is a Petri net
• TASpT, TXSpT, TAJpT, TXJpT are mutually disjoint sets of operator
transitions called AND-split, XOR-split, AND-join and XOR-join
transitions respectively
• j p∈P : •p = ∅
preset (called the source place, usually denoted by i)
• j p∈P : p• = ∅
postset (called the sink place, usually denoted by o)
• ∀t∈T:|•t|N0∧|t•|N0, i.e. all transitions have a non-empty preset as
well as a non-empty postset.
330
331
332
333
fgj =1, i.e. there exists exactly one place with empty
334
335
fgj =1, i.e. there exists exactly one place with empty
336
337
338
Fig. 3 shows an example of a workflow net that uses all four types of
operator transitions. The corresponding decorations are shown on the
339
Fig. 2. A simple Petri net.
4
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 6

340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
transition symbol of TAS= register;sign
TAJ= process;archive
f
a point after which sendform and evaluate can be carried out
simultaneously (or in an arbitrary order); evaluate itself signifies that
adecisionistobemadebetweentwopaths.NotethatinaclassicalPetri
net an XOR-join is represented as a place that has more than one preset
element, e.g. formok and acceptable are such places. Analogously, in a
classical Petri net places with more than one postset element are
equivalenttoXOR-splits.Wewillincludebothtypes ofXORrepresenta-
tions in our workflow nets and consider them as equivalent. However,
wewillrefertotheuseofdecoratedtransitionsforXOR-splitsand–joins
as explicit representations and the use of places for these operators as
implicit presentations. Atthis point, it is important to notethat XOR and
AND operators form the core of the routing logic that is supported by
most process modeling popular languages, e.g. BPMN, UML activity
Diagrams, and EPCs.
A workflow net W=(P,T,F,TAS,TXS,TAJ,TXJ) can always be uniquely
extended by adding a transition t connecting the sink place o with the
source place i. We call t*the short-circuit transition of W. Analogously,
we call the net W*that is represented by the tuple P;T∪ t?
ðo;t?Þ;ðt?;iÞ
W is not a workflow net according to the above definition, because it
has neither a source nor a sink place. We will use it mainly for
technical reasons.
For technical reasons too, we need to be able to reason about the
prefix or precursor of a workflow net, as it exists at intermediate stages
of an interactive modeling session. For this reason, we introduce a
weakenedversionoftheabovedefinitionofaworkflownet.Specifically,
wecallN=(P,T,F,TAS,TXS,TAJ,TXJ)anoperator-extendednetinwhich(P,T,
F)isaPetrinet.Aswillbeshown,thisdefinitionwillallowustohighlight
operator transitions within incomplete workflow nets. Note that the
class of workflownetsisa subsetof theclass of operator-extendednets,
i.e. each workflow net is an operator-extended net.
Finally, we will assume that all operator transitions are used in a
sensible way. Specifically, each split transition has a non-singleton
postset, each join transition has a non-singleton preset, and each non-
operator transition has a singleton preset and postset. Formally, we
call a given operator-extended net N=(P,T,F,TAS,TXS,TAJ,TXJ) operator-
normalized, if the following holds:
fg, TXS= evaluate;check
fg,
g and TXJ=∅. Here, register is used to represent
356
357
358
359
360
361
362
363
364
f g;F∪
ð
fg;TAS;TXS;TAJ;TXJÞ the short-circuited net of W. Note that
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
• ∀t∈TAS∪TXS:|t•|N1∧|•t|≤1
• ∀t∈TAJ∪TXJ:|•t|N1∧|t•|≤1
• ∀t∈T∖(TAS∪TAJ∪TXS∪TXJ):|t•|≤1∧|•t|≤1
380
381
382
383
We are now able to more formally express the notion of matching
operator pairs that we aim to highlight as a way to improve the
comprehension of the overall model. Our definition is a generalization
of the concept of PT/TP-handles [18]:
In an operator-extended net N=(P,T,F,TAS,TXS,TAJ,TXJ), a pair of
nodes (n1,nk)∈(P∪T)×(P∪T) is called a matching operator pair if
384
385
386
387
388
389
• n1∈TASand nk∈TAJor n1∈P∪TXSand nk∈P∪TXJ
• there are at least two elementary paths π1and π2leading from n1to
nkwith αðπ1Þ∩αðπ2Þ = n1;nk
The intuition behind this definition is that if two or more different
paths connect two nodes then these nodes signify the start and end of
a noteworthy sub-component of the overall routing logic. Note that
the notion of matching operator pairs is not restricted to nets that are
completely or even highly block-structured. In the next section, we
will focus on the implementation of the idea to highlight matching
operator pairs.
390
??:
391
392
393
394
395
396
397
398
3.2. Implementation in WoPeD
399
WoPeD2is a Java-based open source tool supporting the modeling
of plain Petri nets as well as that of workflow nets. Several
publications have accompanied the emerging development, e.g.
[13,14]. In the most recent release, the highlighting of matching
operator pairs is supported as a switchable option. If enabled, the
current editor content is constantly monitored for user-inflicted
changes by executing a detection algorithm assigning each matching
operator pair a distinguishable color from a predefined palette.
Fig. 4 shows an operator-normalized workflow net with a total of
four matching operator pairs:
400
401
402
403
404
405
406
407
408
409
• (t1,t6) is an AND-split/AND-join pair (red)
• (t2,t3) is an explicit XOR-split/explicit XOR-join pair (yellow)
• (p5,t5) is an implicit XOR-split/explicit XOR-join pair (green)
• (p11,p7) is an implicit XOR-split/ implicit XOR-join pair (magenta).
410
411
412
413
Note that for determining the colors it is not necessary that the
editor contains a workflow net; an operator-extended net will be
sufficient. For simplicity and without loss of generality, we addition-
ally assume that nets are operator-normalized, i.e. all operators have a
“sensible” branching context. Under this assumption, the set of
414
415
416
417
2http://www.woped.org/.
Fig. 3. A workflow net.
5
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 7

418
419
candidate pairs to be considered as potentially matching operators can
be restricted to pairs (x,y) where either (1) x is an XOR-split (or a place
witha non-singleton postset) and y anXOR-join (or a place witha non-
singleton preset) or (2) x an AND-split and y an AND-join. Additionally,
ifthecurrenteditorcontentisdetectedtoconformtothepropertiesofa
workflownet,theshort-circuitedversionofthenetisconsideredtofind
additional pairs of matching operators. For example, (p11,p7) in Fig. 4
can only be detected by considering the short-circuited version of the
net. Note that as long as the net under construction is not a workflow
net,thedetectionofmatchingoperatorpairsisnotnecessarilycomplete
yet all detected pairs are indeed matches.
Each candidate pair of nodes is checked for being a matching
operator pair by applying the Ford and Fulkerson max-flow-min-cut
algorithm [19]. This algorithm can be used to determine the
maximum flow in a flow network between its source and sink. Our
implementation applies this algorithm in the context of an operator-
extended net by considering for each candidate pair its first (split)
element as a source and its second (join) element as a sink.
Furthermore, the maximum capacity for each edge is set to 1 and
the direction of each edge determines the flow direction. If the max-
flow-min-cut algorithm establishes that the maximum flow between
the two elements of a candidate pair exceeds 1 then this indicates a
minimum of two different elementary paths between them. In other
words, in such a case a matching operator pair is found. In the
example of Fig. 4, the maximum flow between p5and t5is exactly two
since there is a capacity of 1 along the path 〈p5;t8;p6;t5〉 and an
additional capacity of 1 along path 〈p5;t9;p7;t11;p11;t5〉.
Our implementation of the max-flow-min-cut algorithm is derived
from the one introduced in [29], with the modification to select nodes
based on breadth-first-search [15]. With this modification, our
algorithm has a complexity ofO(|D∪D′||A|2). The complete algorithm
executes the max-flow-min-cut algorithm |(P∪T)×(P∪T)| times and
therefore runs in polynomial time with O(|(P∪T)|3|F|2).
At this point, it should be considered that the membership of a
node as an element of a matching operator pair is not exclusive. Pairs
of matching operators may exist that overlap, in the sense that they
share common nodes. Since only a single color can be assigned to each
node at a time, it appears sensible to color all overlapping pairs of
matching operators in the same color. To establish a common
highlighting of overlapping operator pairs, a “clustering” is imple-
mented as shown in Algorithm 1.
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
Algorithm 1. Computation of matching operator clusters
460
461
462
Let OpCpP P ∪T
representation of a matching operator pair (i.e. each pair (x,y) is
contained as a set {x,y})
ðÞ be a set of sets where each element is the set
463
while ∃op1,op2∈OpC:op1≠op2∧op1∩op2≠∅ do
464
op:=op1∪op2
OpC: = Op op1;op2
465
fg∪ op
fg
466
end while
467
return OpC
468
469
To illustrate this algorithm, we start with the example set of
matching operator pairs (t1, t6), (t2, t3), (p5, t5), (t8, t6), (p11, p7) for the
workflow net in Fig. 5. The conversion to a set of sets yields:
OpC=
t1;t6
f
ation,
t1;t6
f
separate elements have been removed and the combined set is added,
no overlapping elements are left in OpC. Each of the remaining
elements is OpCis now a cluster of the underlying net. All elements in a
cluster will be assigned with the same color. As can be seen in Fig. 5,
indeed, t1, t6, and t8are colored the same. The color palette itself can
be created via a settings dialog and configured by the user with
individually chosen color values, as can be seen too in Fig. 5.
Note how this workflow net in Fig. 5 is a slightly extended version
of the block-structured net that is shown in Fig. 4: Transition t8is
turned into an AND-split and an additional place p15is added between
transitions t8and t6. As such, this example also illustrates how the
coloring approach is applicable to both block-structured and non-
block-structured nets.
470
471
472
473
474
475
476
477
478
479
g; t2;t3
t8;t6
f
fg; p5;t5
g are combined into
fg; t8;t6
fg; p11;p7
fg
t1;t6;t8
f
fg. In the first iter-
g. After the
g and
480
481
482
483
484
485
486
4. Research method
487
488
We set up an experiment to investigate the effectiveness of our
proposed highlighting approach by testing the hypotheses of
Section 2. Section 4.1 presents our research design and a description
of how the experiment was conducted. Section 4.2 summarizes the
results of the experiment.
489
490
491
Fig. 4. Highlighted matching operator pairs.
6
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 8

492
4.1. Research design and conduct
493
To design the experiment, we have been following the recom-
mendations given in [32,68]. This section describes the subjects,
objects and selected variables of our experiment, introduces our
hypotheses, and presents the instrumentation and data collection
procedure.
The subjects are 62 experienced modelers both from the industry
and academia (experts) and 41 students following a graduate course
on Business Process Management at Eindhoven University of
Technology (novices). The professionals were recruited from two
Dutch consultancy organizations and three international research
groups with a special focus on Business Process Management. All
participated voluntarily. The central object that is used in the
experiment is the process model that was introduced in Fig. 1. Its
twovariants– withandwithouthighlighting– wereusedtorepresent
two levels of our primary factor of interest, i.e. the highlighting of
matching operator pairs. Furthermore, we recorded the factor
expertise on two levels: expert and novice. In the experiment, two
response variables were used. First, the number of correct answers to a
set of closed questions was used as the indicator of understanding
accuracy. The time that was taken by the respondents to answer this
set questions is the second response variable, indicating the
understanding speed.
To investigate the hypotheses of Section 2, an on-line instrument
was developed to conduct a self-administered experiment. The
instrument was developed with PHP 5.1.4 and JavaScript and ran on
an Apache Web Server hosted at Eindhoven University of Technol-
ogy. The instrument leads the participant through three successive
parts:
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
1. Introduction: In this first part, the purpose of the experiment is
explained as well as the expected effort to complete it. Further-
more, an explanation of the workflow net notation is provided.
Finally, it is explained that participants are expected to answer the
questions accurately and as fast as possible.
522
523
524
525
526
2. Demographic information: The participant is requested to provide
informationonherbackground, experiencewithprocessmodeling,
and familiarity with the workflow net notation.
3. Experiment: Depending on a random draw, the respondent is
assigned with an equal probability to either the model with or
without highlighting support. In total, a set of nine questions is
provided one after another to the respondent. Questions could not
be skipped. At all times, a legend with an explanation on the
workflow net notation is visible.
527
528
529
530
531
532
533
534
535
The screenshot in Fig. 6 gives an impression of the look and feel of
the instrument.
The closed questions from one of our previous experiments [42]
were used in this experiment. An analysis of these questions and the
data obtained in that work provided a value of 0.675 for Cronbach's
alpha. This can be considered as an acceptable indication for internal
consistency, for example when compared with other empirical
research in the context of business process oriented research [28].
The questions cover issues with respect to:
536
537
538
539
540
541
542
543
544
• concurrency, e.g. “Can tasks A and B be executed at the same time for
a case?”
• exclusiveness, e.g. “Can tasks A and B both be executed for the same
case?”
• order, e.g. “If task A is executed for a case, must thenalways taskB be
executed for the same case?”
• repetition, e.g.“Can task A be executed more than once for the same
case?”
545
546
547
548
549
550
551
552
For all questions, the respondents were offered the answer
categories ‘yes’, ‘no’ and ‘do not know’.
553
554
4.2. Results
555
556
Beforetesting thehypotheses,thequalityofthedatawasanalyzed.
Fromthe total of 103 answer sets thatwere provided for the questions
on the models (both with and without highlighting), 26 were
557
Fig. 5. A settings dialog for the color palette. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
7
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 9

558
incomplete. Although the tool did not allow any questions to be
skipped, people always had the opportunity to stop the experiment at
any time by simply closing their browser. All data being part of an
incomplete set of answers was removed. Furthermore, in line with
[22], quality checking was applied to the data. We expected expert
modelers to be able to at least correctly answer 5 out of 9 questions
(55%) and novices at least 4 out of 9 questions (44%). Results below
these thresholds were considered to reflect a serious lack of
knowledge and/or commitment. As a result, 7 additional answer
sets had to be removed. The latter is a relatively small number, which
is an indication for the overall high quality of the raw data. In total, 70
answer sets were taken into account for the data analysis (=103−
(26+7)).
559
560
561
562
563
564
565
566
567
568
569
570
571
4.2.1. General effect of highlighting
The 70 answer sets were explored for the response variables of
interest, understanding accuracy and understanding speed. The subjects
572
573
574
on average provided 7.21 correct answers for the total of 9 questions
(80%) with a minimum of 5 and a maximum of 9. On average, it took
respondents a little over 10 min to answer the questions, with a
minimum of approximately 4 min and a maximum of approximately
34 min. Application of the Kolmogorov–Smirnov test [59] indicated
that the distributions for both variables differ significantly from the
normal distribution, which may be due to the limited size of the data
set. Box plots that show the distribution of the response variables,
differentiating between the factor levels with and without highlight-
ing, are provided in Fig. 7.
Inspection of the box plots does not point at understanding speed
being affected strongly by highlighting, i.e. the medians and
distributions are quite similar. The understanding accuracy for subjects
provides a somewhat other picture, with the number of correct
answers being slightly higher for those subjects that used the process
models with highlighting support for answering the questions. A
corresponding non-parametric Mann–Whitney test [59] confirms
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
Fig. 6. Screen shot of the on-line instrument.
Fig. 7. Boxplots for understanding accuracy (number of correct answers) and understanding speed (time).
8
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 10

591
592
these observations. This type of test is appropriate in the light of the
non-normality of the data. The Mann–Whitney test is applicable to
experiments that consist of one factor and two treatments with a
completely randomized design [68], asis the casehere. Table 1 reports
thesignificancelevels.Accordingly,wefindsupportfor Hypothesis H1
(i.e. there is a positive impact on understanding accuracy), but not for
Hypothesis H2 (i.e. there is no positive impact on understanding
speed).
593
594
595
596
597
598
599
600
4.2.2. Effect of coloring for novices and experts
A more detailed data analysis that distinguishes between novices
and experts is provided in Table 2.
From the table, various differences can be distinguished between
the analyzed statistics. For example, experts on average perform
better than novices: They have higher numbers of correct answers
and are faster in responding to the questions. Also notable is that
novices on average correctly answer 7.08 questions when provided
with highlighting support but only 5.92 without that support, which
would be in line with Hypothesis H3. To examine all the hypotheses
we stated, we used the non-parametric Mann–Whitney test sepa-
rately of the novice sub-sample and the expert sub-sample. The
results are shown in Table 3.
As can be seen, a significant difference between the availability of
highlighting support and the lack thereof exists and relates to the
understanding accuracy (number of correct answers) for novices.
Indeed, the difference we mentioned between the 7.08 and 5.92
correct answers on average is not likely to be a matter of chance. In
other words, whether a novice model reader is considering a process
model with or without highlighting will make a significant difference
in terms of her understanding accuracy. In fact, that accuracy is then
higher. The understanding speed for novices, however, does not differ
significantly. Furthermore, highlighting is not a distinguishing factor
for the performance of the experts.
In summary,wefind support for thegeneral benefit of highlighting
for the understanding accuracy (H1), in particular for novices (H3).
The effect is not significant for experts though (H5). Effects on
understanding speed appear to be negligible (H2, H4, and H6).
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
5. Discussions and implications
628
629
In this section we discuss the results and the implications of our
research.
630
5.1. Discussion of results
631
632
The results of the experiment have to be interpreted in terms of
significance and strength of the noted effect. Furthermore, our
distinction between novices and experts has to be reflected upon.
Personal factors have been identified as factors of process model
understanding in prior research [42,44,53]. In this experiment we
considered a classification of novices and experts based on partici-
pants being students or professionals (either from academia or
industry). In general, the distinction between novices and experts is
not so straightforward and related to notions that are difficult to
measure, like “the amount and complexity of knowledge gained
through extensive experience of activities in a domain” [16] or “the
result of acquiring […] vast amounts of knowledge and the ability to
perform pattern-based retrieval” [10]. Furthermore, it is important to
notice that task-specific experience is often a better predictor of
performance than expertise in general [6]. The fact that the mean
values in our experiment are conclusive supports the appropriateness
of our classification: experts without highlighting are still faster and
more accurate than novices with highlighting in Table 2.
The experiment showed that the highlighting was of greatest
benefitto theaccuracyof novices, suchthatitwasthesinglesignificant
effect. This observation is in line with research that establishes
modeler expertise as a critical issue for process modeling projects [3].
Petre observed in her research on secondary notation that novices
tend to lack reading and search strategies which result from modeling
experience and extensive learning [51]. Syntax highlighting in our
experimentis a significant aid for novices to read the models and their
control flow semantics. Therefore, it is no surprise that their
performance in terms of accuracy is improved. Experts, in turn, can
much easier identify patterns of matching operator transitions, a skill
that is also referred to as perceptual expertise [25]. Accordingly, the
highlighting helps them to identify patterns they already know. As a
consequence, the performance increaseis too small to be significant in
our experiment.
It is arguable that the effect of highlighting on performance of both
experts and novices might have been stronger if the models had been
more complex. It is well known from prior research that more
complex models are more difficult to understand [39,42]. Several
metrics have been proposed to measure different dimensions of
complexity of a process model, e.g. in [1,8,9,36,39,45,48,49,64]. The
models we used in the experiment are fairly structured such that a
split operator most often has a direct join counterpart. Such
structured models are rather easy to understand for experts. The
highlighting effect might have been more effective also for experts if
the models had been more unstructured. The reader may recall that,
indeed, the identification of matching operator pairs is also possible in
unstructured nets. Additionally, it might be argued that models need
to be much larger before highlighting starts to have a significant effect
on experts' performance.
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
Table 1
Two-tailed Mann–Whitney test for Hypotheses H1 and H2.
t1:1
t1:2
t1:3
Hypothesis p-valueSupport
t1:4
t1:5
H1 Highlighting→Accuracy
H2 Highlighting→Speed
0.049
0.583
✓
∅
Table 2
Descriptive statistics for understanding accuracy (number of correct answers) and
understanding speed (time).
t2:1
t2:2
t2:3
NovicesExperts
t2:4
t2:5
With
highlighting
Without
highlighting
With
highlighting
Without
highlighting Statistics
t2:6
t2:7
Number of
Correct
answers
N
Mean
1313
5.92
1826
7.46 7.08 7.89
t2:8
t2:9
t2:10
t2:11
t2:12
Median
Std. Dev.
Skewness
Kurtosis
N
7.00
1.188
−.524
−.105
13
6.00
.954
.854
.221
13
8.00
1.231
−1.041
0.289
18
7.50
1.140
−.161
−.697
26Time
(minutes)
t2:13
t2:14
t2:15
t2:16
t2:17
Mean
Median
Std. Dev.
Skewness
Kurtosis
102.680
79.333
795.681
2.587
7.621
115.449
99.167
486.906
.694
.224
98.796
87.083
445.450
1.471
1.606
96.231
81.917
536.402
2.251
4.683
Table 3
Two-tailed Mann–Whitney test for Hypotheses H3 to H6.
t3:1
t3:2
t3:3
Hypothesis p-ValueSupport
t3:4
t3:5
t3:6
t3:7
H3 Highlighting→Accuracy of Novices
H4 Highlighting→Speed of Novices
H5 Highlighting→Accuracy of Experts
H6 Highlighting→Speed of Experts
0.017
0.139
0.184
0.535
✓
∅
∅
∅
9
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 11

679
5.2. Implications
680
681
The experiment and the results have implications for research and
practice. With this work we have shown the potential of secondary
notation to improve process model understanding. We have focused
on the highlighting of matching operator pairs. Further research is
needed to investigate the effect of other types of secondary notation.
In particular, techniques from automatic graph drawing have been
discussed regarding their support of understanding, e.g. in [67]. A
dedicated discussion of automatic layouting of process models and
their benefit to comprehension is missing so far. For such a research
endeavor it is important to consider modeling expertise and its
interplay with secondary notation. While process modeling expertise
has been considered by some studies on process model comprehen-
sion [42,44,53], there is a need for further research on establishing a
more detailed foundation for this notion.
In this article, we also demonstrated the feasibility of automatic
highlighting both by providing a formalization and by an implemen-
tation within the Petri net modeling tool WoPeD. The general
concepts of our approach can be easily extended to other activity-
oriented process modeling languages (e.g. UML Activity Dia-
grams [17], EPCs [62], or BPMN [12]). Even for those languages that
do not directly build on token passing semantics, graph parsing
techniques can be used to define colors based on so-called single-
entry-single-exit components [31,66]. Therefore, our highlighting
approach can be directly included in industry process modeling tools
to improve process model comprehension.
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
6. Conclusion
706
707
This paper addresses the problem of how process model users can
be better supported in the sense-making of a model. We adapted the
concept of syntax highlighting from software engineering and applied
it to workflow nets, a formalism that is frequently used for business
process modeling. Our contribution is a theoretical discussion of the
benefits of such highlighting, a formalization of the highlighting
problem along with a prototypical implementation, and a thorough
experimental study on the effects of highlighting on comprehension
performance. Due to the structural and semantic similarities between
workflownetsandotherprocessmodelinglanguages,theresultsfrom
this research can be easily applied in other process modeling tools.
Our future work is in line with some of the open issues we already
noted. We plan to conduct further experiments to study the
interaction of process model complexity with the effects of highlight-
ing. The assumption for this work could be that highlighting benefits
increase with an increase in model complexity.
Beyond this, there are other types of secondary notation that we
did not study in this research. Specifically, we aim to investigate to
which degree the comprehension of a process model can be improved
by a good layout. Prior work in the area of class diagrams suggest that
comprehension would deteriorate with bad layout [67]. The challenge
in this context will be, among others, to operationalize the notion of a
good layout for process models.
Furthermore, we want to explore the spectrum of visual
parameters [47] that may be utilized to enhance process model
comprehension. More specifically, many popular process modeling
notations include a range of symbols with a fixed graphical form
without any notable consideration of their cognitive discriminability.
It would be interesting to work on a “make over” of such notations to
improve their use; the developers of the YAWL modeling language
[63], for example, have already expressed their willingness to
cooperate in such an endeavor.
Finally, we also have an interest in developing guidelines for
process modelers with respect to the way they structure their models.
The work on seven process modeling guidelines that we presented in
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
[43] can be seen as a first step in that direction, which offers
considerable potential for extension.
743
References
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
[1] E.R. Aguilar, F. García, F. Ruiz, M. Piattini, An exploratory experiment to validate
measures for business process models, First International Conference on Research
Challenges in Information Science (RCIS), 2007, pp. 271–280.
[2] J. Anderson, ACT: a simple theory of complex cognition, The American
Psychologist 51 (4) (1996) 355–365.
[3] W. Bandara, Factors and measures of business process modelling: model building
through a multiple case study, European Journal of Information Systems 14
(2005) 347–360.
[4] J. Bertin, Semiology of Graphics, University of Wisconsin Press, 1983.
[5] A.F. Blackwell, Ten years of cognitive dimensions in visual languages and
computing: guest editor's introduction to special issue, Journal of Visual Language
and Computing 17 (4) (2006) 285–287.
[6] S. Bonner, N. Pennington, Cognitive processes and knowledge as determinants of
auditor expertise, Journal of Accounting Literature 10 (1) (1991) 1–50.
[7] A. Burton-Jones, Y. Wand, R. Weber, Guidelines for empirical evaluations of
conceptual modeling grammars, Journal of the Association for Information
Systems 10 (6) (2009) 495–532.
[8] G. Canfora, F. García, M. Piattini, F. Ruiz, C. Visaggio, A family of experiments to
validate metrics for software process models, The Journal of Systems and Software
77 (2) (2005) 113–129.
[9] J. Cardoso, Workflow Handbook 2005, chap, Evaluating Workflows and Web
Process Complexity, Future Strategies, Inc, Lighthouse Point, FL, USA, 2005,
pp. 284–290.
[10] W. Chase, H. Simon, The mind's eye in chess, Visual Information Processing 215
(1973) 215–281.
[11] I. Davies, P. Green, M. Rosemann, M. Indulska, S. Gallo, How do practitioners use
conceptual modeling in practice? Data & Knowledge Engineering 58 (3) (2006)
358–380.
[12] R.M. Dijkman, M. Dumas, C. Ouyang, Semantics and analysis of business process
models in BPMN, Information and Software Technology 50 (12) (2008)
1281–1294.
[13] A. Eckleder, T. Freytag, WoPeD 2.0 goes BPEL 2.0, in: N. Lohmann, K. Wolf (eds.),
Proceedings of the 15th German Workshop on Algorithms and Tools for Petri
Nets, AWPN 2008, Rostock, Germany, September 26–27, 2008, vol. 380 of CEUR
Workshop Proceedings, CEUR-WS.org, 2008, pp. 75–80
[14] A. Eckleder, T. Freytag, J. Mendling, H. A. Reijers, Realtime detection and coloring
of matching operator nodes in workflow nets, in: T. Freytag, A. Eckleder (eds.),
16th German Workshop on Algorithms and Tools for Petri Nets, AWPN 2009,
Karlsruhe, Germany, September 25, 2009, Proceedings, vol. 501 of CEUR
Workshop Proceedings, CEUR-WS.org, 2009, pp. 56–61.
[15] J. Edmonds, R.M. Karp, Theoretical improvements in algorithmic efficiency for
network flow problems, Journal of the ACM 19 (2) (1972) 248–264.
[16] K. Ericsson, A. Lehmann, Expert and exceptional performance: evidence of
maximal adaptation to task constraints, Annual Review of Psychology 47 (1)
(1996) 273–305.
[17] R. Eshuis, R. Wieringa, Tool support for verifying uml activity diagrams, IEEE
Transactions on Software Engineering 30 (7) (2004) 437–447.
[18] J. Esparza, M. Silva, Circuits, handles, bridges and nets, in: G. Rozenberg (Ed.),
Advances in Petri Nets 1990, vol. 483, 1990, pp. 210–242.
[19] L.R. Ford, D.R. Fulkerson, Maximal flow through a network, Canadian Journal of
Mathematics 8 (3) (1956) 399–404.
[20] P. Frederiks, T. Weide, Information modeling: the process and the required
competencies of its participants, Data & Knowledge Engineering 58 (1) (2006)
4–20.
[21] A. Gemino, Y. Wand, A framework for empirical evaluation of conceptual
modeling techniques, Requirements Engineering 9 (4) (2004) 248–260.
[22] M. Genero, G. Poels, M. Piattini, Defining and validating metrics for assessing the
understandability of entity-relationship diagrams, Data & Knowledge Engineering
64 (3) (2008) 534–557.
[23] T. Green, Cognitive dimensions of notations, in: A. Sutcliffe, L. Macaulay (Eds.),
People and Computers V: Proceedings of the Fifth Conference of the British
Computer Society Human-Computer Interaction Specialist Group, 1989,
pp. 443–460.
[24] T. Green, M. Petre, Usability analysis of visual programming environments: a
‘cognitive dimensions’ framework, Journal of Visual Languages and Computing 7
(2) (1996) 131–174.
[25] T. Green, W. Ribarsky, B. Fisher, Building and applying a human cognition model
for visual analytics, Information Visualization 8 (1) (2009) 1–13.
[26] A. Guceglioglu, O. Demirors, Using software quality characteristics to measure
business process quality, Proc. BPM'05, Springer, 2005, pp. 374–379.
[27] D. Harel, On visual formalisms, Communications of the ACM 31 (5) (1988)
514–530.
[28] R. Hung, Business process management as competitive advantage: a review and
empirical study, Total Quality Management & Business Excellence 17 (1) (2006)
21–40.
[29] T. Ihringer, Diskrete Mathematik, Heldermann, 2002.
[30] M. Indulska, J. Recker, M. Rosemann, P. Green, Process modeling: current issues
and future challenges, Advanced Information Systems Engineering-CAiSE, 2009,
pp. 501–514.
10
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013

Page 12

823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
[31] R. Johnson, D. Pearson, K. Pingali, The program structure tree: computing control
regions in linear time, Proceedings of the ACM SIGPLAN'94 Conference on
Programming Language Design and Implementation (PLDI), Orlando, Florida, June
20–24, 1994, SIGPLAN Notices, Vol. 29 (6), 1994, pp. 171–185.
[32] N. Juristo, A.M. Moreno, Basics of Software Engineering Experimentation,
Springer, 2001.
[33] N. Kock, J. Verville, A. Danesh-Pajou, D. DeLuca, Communication flow orientation
in business process modeling and its effect on redesign success: results from a
field study, Decision Support Systems 46 (2) (2009) 562–575.
[34] M. Koubarakis, D. Plexousakis, A formal framework for business process
modelling and design, Information Systems 27 (5) (2002) 299–319.
[35] J. Larkin, H. Simon, Why a diagram is (sometimes) worth ten thousand words**,
Cognitive Science 11 (1) (1987) 65–100.
[36] G. Lee, J.-M. Yoon, An empirical study on the complexity metrics of Petri nets,
Microelectronics Reliability 32 (3) (1992) 323–329.
[37] G. Lohse, A cognitive model for understanding graphical perception, Human-
Computer Interaction 8 (4) (1993) 353–388.
[38] J. Mackinlay, Automating the design of graphical presentations of relational
information, ACM Transactions on Graphics 5 (2) (1986) 110–141.
[39] J. Mendling, Metrics for process models, Empirical Foundations of Verification,
Error Prediction, and Guidelines for Correctness, vol.6 of Lecture Notes inBusiness
Information Processing, Springer, 2008.
[40] J. Mendling, Empirical studies in process model verification, LNCS Transactions on
Petri Nets and Other Models of Concurrency II, Special Issue on Concurrency in
Process-Aware Information Systems 2 (2009) 208–224.
[41] J. Mendling, G. Neumann, W. Aalst, Understanding the occurrence of errors in
process models based on metrics, in: R. Meersman, Z. Tari (Eds.), OTM Conference
2007, Proceedings, Part I, vol. 4803 of Lecture Notes in Computer Science,
Springer, 2007, pp. 113–130.
[42] J. Mendling, H.A. Reijers, J. Cardoso, What makes process models understandable?
Proc. BPM'07, 2007, pp. 48–63.
[43] J. Mendling, H.A. Reijers, W.M.P. van der Aalst, Seven process modeling guidelines
(7PMG), Information and Software Technology 52 (2) (2009) 127–136.
[44] J. Mendling, M. Strembeck, Influence factors of understanding business process
models, in: W. Abramowicz, D. Fensel (Eds.), Proc. of the 11th International
Conference on Business Information Systems (BIS 2008), vol.7 of Lecture Notes in
Business Information Processing, Springer-Verlag, 2008, p. 142153.
[45] J. Mendling, H.M.W. Verbeek, B.F. van Dongen, W.M.P. van der Aalst, G. Neumann,
Detection and prediction of errors in EPCs of the SAP reference model, Data &
Knowledge Engineering 64 (1) (2008) 312–329.
[46] C. Minkowitz, Formal process modelling, Informationand Software Technology 35
(11) (1993) 659–667.
[47] D. Moody, The “physics” of notations: toward a scientific basis for constructing
visual notations in software engineering, IEEE Transactions on Software
Engineering 35 (6) (2009) 756–779.
[48] S. Morasca, Measuring attributes of concurrent software specifications in Petri
nets, METRICS '99: Proceedings of the 6th International Symposium on Software
Metrics, IEEE Computer Society, Washington, DC, USA, 1999, pp. 100–110.
[49] M. Nissen, Redesigning reengineering through measurement-driven inference,
MIS Quarterly 22 (4) (1998) 509–534.
[50] D. Parnas, On the criteria for decomposing systems into modules, Communica-
tions of the ACM 15 (12) (1972) 1053–1058.
[51] M. Petre, Why looking isn't always seeing: readership skills and graphical
programming, Communications of the ACM 38 (6) (1995) 33–44.
[52] M. Petre, Cognitive dimensions ‘beyond the notation’, Journal of Visual Languages
and Computing 17 (4) (2006) 292–301.
[53] J. Recker, A. Dreiling, Does it matter which process modelling language we teach
or use? An experimental study on understanding process modelling languages
without formal education, in: M. Toleman, A. Cater-Steel, D. Roberts (Eds.), 18th
Australasian Conference on Information Systems, The University of Southern
Queensland, Toowoomba, Australia, 2007, pp. 356–366.
[54] H.A. Reijers, J. Mendling, Modularity in process models: review and effects,
Proceedings of BPM 2008, Springer, 2008, pp. 20–35.
[55] H.A. Reijers, R.S. Mans, R.A. van der Toorn, Improved model management with
aggregated business process models, Data & Knowledge Engineering 68 (2)
(2009) 221–243.
[56] M. Rosemann, Potential pitfalls of process modeling: part a, Business Process
Management Journal 12 (2) (2006) 249–254.
[57] A.W. Scheer, ARIS business process modelling, Springer Verlag, 2000.
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
[58] K. Siau, Informational and computational equivalence in comparing information
modeling methods, Journal of Database Management 15 (1) (2004) 73–86.
[59] S. Siegel, N.J. Castellan, Nonparametric Statistics for the Behavorial Sciences, 2nd
ed, McGraw Hill, 1988.
[60] W.M.P. van der Aalst, K.M. van Hee, Workflow Management: Models, Methods,
and Systems, The MIT press, 2004.
[61] W.M.P. van der Aalst, The application of Petri nets to workflow management,
Journal of Circuits Systems and Computers 8 (1) (1998) 21–66.
[62] W.M.P. van der Aalst, Formalization and verification of event-driven process
chains, Information and Software Technology 41 (10) (1999) 639–650.
[63] W.M.P. van der Aalst, A.H.M. ter Hofstede, YAWL: Yet Another Workflow
Language, Information Systems 30 (4) (2005) 245–275.
[64] I. Vanderfeesten, H.A. Reijers, J. Mendling, W.M.P. van der Aalst, J. Cardoso, On a
quest for good process models: the cross-connectivity metric, Lecture Notes in
Computer Science 5074 (2008) 480–494.
[65] I. Vanderfeesten, H.A. Reijers, W.M.P. van der Aalst, Evaluating workflow process
designs using cohesion and coupling metrics, Computers in Industry 59 (5)
(2008) 420–437.
[66] J. Vanhatalo, H. Völzer, J. Koehler, The refined process structure tree, Data &
Knowledge Engineering 68 (9) (2009) 793–818.
[67] C. Ware, H.C. Purchase, L. Colpoys, M. McGill, Cognitive measurements of graph
aesthetics, Information Visualization 1 (2) (2002) 103–110.
[68] C. Wohlin, R. Runeson, M. Halst, M. Ohlsson, B. Regnell, A. Wesslen, Experimen-
tation in Software Engineering: An Introduction, Kluwer, 2000.
Hajo Reijers is an associate professor at the School of Industrial Engineering of
Eindhoven University of Technology. He received a PhD in Computer Science from
Eindhoven University of Technology, while being a manager with Deloitte Consulting.
His research interests are in business process modeling, workflow management
technology, and discrete event simulation. He published on these and other topics in
the Journal of Management Information Systems, Information Systems, Journal of
Information Technology, Data and Knowledge Engineering, Organization Studies, and
other scholarly journals. He is an associate editor of two international journals and
founder of the Dutch BPM-Forum, a platform for knowledge exchange between
industry and academia related to business process optimization. He was the program
co-chair of the International Conference on Business Process Management 2009.
Thomas Freytag is a professor at the Cooperative State University Baden-Wuerttem-
berg (DHBW) Karlsruhe, Germany. He holds a university diploma in computer science
and a doctorate degree in economic sciences. Between his two academic degrees, he
worked seven years as a software developer and project manager in a large postal
automation company. His teaching and research areas at the university are software
development, system analysis and business process management. Since 2003, he is
supervising the development of WoPeD, an open source tool in the area of Petri-net-
based workflow modeling and analysis. Besides his academic career, he owns a small
consulting company which is embedded into the “Steinbeis” foundation, a large,
government-driven knowledge transfer network.
Jan Mendling is a Junior-Professorat the Institute of Information Systems at Humboldt-
Universität zu Berlin, Germany. His research areas include Business Process Manage-
ment, Conceptual Modelling and Enterprise Systems. He has published more than 100
research papers and articles, among others in ACM Transactions on Software
Engineering and Methodology, Information Systems, Data and Knowledge Engineering,
Formal Aspects of Computing, and Information and Software Technology. He is member
of the editorial board of two international journals. His Ph.D. thesis has won the Heinz-
Zemanek-Award of the Austrian Computer Society and the German Targion-Award for
dissertations inthe area of strategic information management.He is oneof the founders
and chair of the Berlin BPM Community of Practice (http://www.bpmb.de) and
organizer of several academic events on process management. He will be program co-
chair of the International Conference on Business Process Management 2010.
Andreas Eckleder is an external lecturer at the Cooperative State University Baden-
Wuerttemberg (DHBW) Karlsruhe, Germany. His teaching area at the university is
system analysis. He holds a diploma degree in commercial computer science from the
DHBW Karlsruhe. Besides of lecturing, he is a department and project manager at Nero
Development and Services GmbH in Karlsbad, Germany. He has been actively
participating in the development of WoPeD since 2006 and continues to coach and
support students working on the project.
11
H.A. Reijers et al. / Decision Support Systems xxx (2011) xxx–xxx
Please cite this article as: H.A. Reijers, et al., Syntax highlighting in business process models, Decis. Support Syst. (2011), doi:10.1016/j.
dss.2010.12.013