![](bg3.png)
Delayed Decissions
Business Language
Configurable BP Model
1.
X
Automatic
Business
Expert
Answers
3.
Guided Execution
Partial BP Model
4.
Questionnaire
2.
Figure 2: Overview of our contribution.
Book hotel
Select clothes
Pack luggage
0 1 2 3 4 5
select
+
(b) BP Graph (BPMN)
pack
book
+
(a) Gantt Chart
Figure 3: A schedule (a) as a BP Graph (b).
Example 1. Figure 3 (a) shows three activities which
are scheduled to prepare a holiday and they are de-
picted as a Gantt chart (Gantt, 1913). The activ-
ities ’book a hotel’, ’select the clothes’ and ’pre-
pare the luggage’ are considered. In addition, Fig.
3 (b) shows the related BP graph using BPMN.
This graph consists of the following 7 nodes (cf.
Def. 1): <1, start, event>, <2, AND, gateway>,
<3, book, activity>, <4, select, activity>, <5, AND,
gateway>, <6, pack, activity> and <7, end, event>;
which are paired (cf. Def. 1) as follows: (1, 2), (2, 3),
(2, 4), (3, 5), (4, 5), (5, 6), and (6, 7).
2.2 Configurable BP Model
Typically, different BPs (cf. Def. 1), also called
variants, can be performed in scenarios which en-
tail high variability. In most cases these plans share
many commonalities. Hence, these variations can be
combined in a configurable BP model (i.e., a mod-
elling artifact that capture a family of BP models in
an integrated manner) leading to a compact represen-
tation (Rosa et al., 2012; Rosemann and van der Aalst,
2007; La Rosa et al., 2008; van der Aalst et al., 2006).
Generally, configurable BP models allow analysts to
understand what these variations share, what their dif-
ferences are, and why and howthese differences occur
(Rosemann and van der Aalst, 2007).
Configurable BP models are typically created by
hand (1) from scratch, (2) from an existing BP model
by including possible adaptations (Gottschalk et al.,
2008), or (3) by merging some BP models related to
the same or similar goals which already exist (Rosa
et al., 2012; Jimenez-Ramirez et al., 2013). In the
last case, the source BP models need to be compared
and merged, which might result in a tedious, time-
consuming and error-prone process if it is performed
by hand (Rosa et al., 2012). To overcome these prob-
lems, there exist approaches focused on automatically
merging different BP models in a configurable BP
model (Rosa et al., 2010; Rosa et al., 2012).
Configurable BP models can be represented by
configurable BP graphs, which are defined (cf. Def.
2) based on (Rosa et al., 2012).
Definition 2. A Configurable BP Graph CG =
(G, E2I, N2LI) consists of: (1) a graph, G =
(gid, N, Pairs) (cf. Def. 1), (2) a function E2I that
maps each edge e ∈ Pairs to a set of BP graph identi-
fiers (i.e., E2I identifies which branches of CG belong
to each source BP graph which is merged in CG), (3)
a function, N2LI that maps each node n ∈ N to a set
of pairs < gpid, l > where gpid is a BP graph iden-
tifier and l is the label of node n in graph gpid (i.e.,
N2LI identifies which nodes, with the corresponding
label, belong to each graph which is merged in CG).
A configurable BP graph includes configuration
nodes for those points where the BP graphs which
are included differ (cf. Example 2). Therefore, each
branch and node of the configurable BP graph can be
related either to one or more BP graphs. To store these
relations, each branch/node of the configurable BP
graph includes identifiers related to the correspond-
ing BP graph (i.e., E2I function). In addition, nodes
also store the associated label related to each identifier
(i.e., N2LI function).
Example 2. Figure 4 shows 2 graphs which are
merged into a configurableBP model
1
. The first gate-
way in Fig. 4(b) is a configurable node which corre-
sponds to an ’OR’ gateway in the process 1 (it does
not explicitly appear) and an ’AND’ gateway in the
process 2.
2.3 Questionnaires
Questionnaire models (Rosa et al., 2009) are gener-
ally created to support the user during the individual-
1
As there is not ambiguity, some labels are not shown
(i.e., they are the same as in the branch).
AutomaticGenerationofQuestionnairesforManagingConfigurableBPModels
711