Information Technology Reference
In-Depth Information
through model generation. This approach was also taken by most of the other
works which built a similar system [17,22,18,23]. The data structure used by the
approach of the University of Rio de Janeiro [23] was taken from the CREWS
project [15]. The authors argue that it is suited well for this task as a scenario
description corresponds to the description of a process model. Therefore, we also
use the CREWS scenario metamodel as starting point. However, we modified
several parts as, e.g., we explicitly represent connections between the elements
using the class “Flow”. Additionally, we explicitly considered traceability as a
requirement. Thus, attributes relating an object to a sentence or a word are
added to the World Model. The four main elements of our World Model are
Actor, Resource, Action, and Flow. This World Model will be used throughout
all phases of our transformation procedure to capture syntactic and semantic
analysis results. Each phase is allowed to access, modify and add data.
The rest of this section is dedicated to analyzing and discussing the issues
collected in Table 1. We will then seize the developed suggestions and reference
these issues during the description of our transformation approach. Section 3.1
discusses sentence level analysis for finding actions. Section 3.2 investigates text
level analysis for enriching the data stored in the world model. Finally, Sec-
tion 3.3 describes the generation of a BPMN model. While we focus on the
general procedure here, we documented details of all algorithms in [25].
3.1 Sentence Level Analysis
The first step of our transformation procedure is a sentence level analysis. The
extraction procedure consists of the steps that are outlined as a BPMN model
in Figure 2. This overview also shows the different components upon which our
transformation procedure builds and their usage of Data Sources.
The text is processed in several stages. First, a tokenization splits up the text
into individual sentences. The challenge here is to distinguish a period used for
an abbreviation (e.g. M.Sc.) from a period marking the end of a sentence.
Fig. 2.
Structural overview of the steps of the Sentence Level Analysis
