Information Technology Reference
In-Depth Information
Fig. 5.
The claims handling model as generated by our system
modeling languages to BPMN in order to compare them. Table 2 lists character-
istics of the texts and models of our data set. The table captures the following
data: a unique ID, the number of models (M), the number of sentences (n), the
average length of sentences (
∅
l), the size of the models in terms of nodes (
|
N
|
),
gateways (
). All our material is published in [25].
The evaluation results are based on the similarity (sim) between the manually
and automatically created models. We employ the metric of
Graph Edit Distance
.
To compute the Graph Edit Distance, the graph representation of the process
models is analyzed. The labels, attributes, the structural context, and behavior
are compared [32]. Afterwards a greedy graph matching heuristic [33] is employed
|
G
|
), and edges (
|
E
|
Table 2.
Characteristics of the test data set by source (average values)
ID
Source
M Type
n
∅
l
|N|
|G|
|E|
1
HU Berlin
4
academic
10.00 18.14 25.75 6.00 24.50
2
TU Berlin [30]
2
academic
34.00 21.17 71.00 9.50 79.50
3
QUT
8
academic
6.13 18.26 14.88 1.88 16.00
4
TU Eindhoven [31]
1
academic
40.00 18.45 38.00 8.00 37.00
5
Vendor Tutorials
4
industry
9.00 18.20 14.00 2.25 11.50
6
inubit AG
4
industry
11.50 18.38 24.00 4.25 21.25
7
BPM Practicioners
1
industry
7.00
9.71 13.00 1.00
9.00
8
BPMN Prac. Handbook [9]
3
textbook
4.67 17.03 13.00 1.33 14.67
9
BPMN M&R Guide [29]
6
textbook
7.00 20.77 23.83 3.00 23.67
10 FNA - Metrology Processes 14 public sector
6.43 13.95 24.43 3.14 25.93
Total
47
9.19 17.16 23.21 3.38 23.64
