Information Technology Reference
In-Depth Information
4.1.1 Structure Pre-processing (Activity 2.1)
The first step to understand how the viewpoint model parser builds the underlying
model and populates the objects is to understand the concept of a parsing unit. For
EA-Miner a parsing unit is the minimal unit that represents a requirement, for
instance, a sentence, paragraph, set of paragraphs, section in a document.
The parsing unit is used during EA-Miner's parsing to allocate requirements for the
concepts as will be detailed in the following sections, for example grouping a set of
sentences as requirements of a specific viewpoint. The reason why a parsing unit can
vary is because input documents for the tool can have varied structured (e.g., a
general description such as in Fig. 2a, an interview transcript, a legacy use-case
specification, a user manual). Therefore, a sentence is not always a good parsing unit.
For example, in structured legacy specifications, normally there is already a natural
grouping of requirements in the document (e.g., organized by section).
The task of EA-Miner in this pre-processing can require a varied level of
interaction with the requirements engineer. The default parsing unit assumed by EA-
Miner is a sentence. If the user accepts this as the unit then no further pre-processing
interaction is required as the file that comes back from WMATRIX is already tagged
in sentences as shown in Fig. 2b. However, if the user wants a different parsing unit
then the interaction is more active such as:
•
The user is required to define patterns that represent parsing units, so EA-Miner
can try to identify and flag them so that the user can accept or discard them. For
example if one is mining a legacy document where requirements are grouped by
Use Case ID followed by name (e.g., UC001: Register User) a pattern can be
defined for that in a form of a regular expression.
•
The user can also use some features that EA-Miner can offer such as selecting a
chunk of text in the editor and clicking a button such as “tag as parsing unit”.
•
The output of the pre-processing activity is basically the same file shown in Fig. 2b
with extra annotation in the form of parsing units (<parsingUnit> </parsingUnit>).
If the user has selected the default sentence parsing unit, then no extra annotation is
done. After the pre-processing process is complete the user moves on to the
identification activity described below.
4.1.2 Viewpoint Identification, Presentation and Structuring (Activities 2.2
and 3.1)
After clarifying the concept of the parsing unit, now we explain how the viewpoint
model parser builds the underlying model and populates the objects. For the examples
in this section and the following ones we consider the parsing unit to be a sentence
as it is more appropriate for the type of document shown in Fig. 2a. As shown in
Fig. 2b each sentence of the input file is tagged with (<s> </s>) tags containing
groups of tagged words (<w pos=”value” sem=”value”> word </w>). The basic
algorithm of the parser is as follows:
•
For each sentence in the tagged file, the parser identifies each sentence as a
requirement;
•
For each word contained in a sentence the parser identifies the viewpoints (and
their related requirements) and early aspects (and their related requirements)
based on some rules (e.g., nouns are candidates for viewpoints);
Search WWH ::
Custom Search