Information Technology Reference
In-Depth Information
scattered and tangled across the documents. Moreover, the fact that AORE is a novel
approach complicates the analysis since many system analysts do not have good
understanding of early aspects.
This is where the EA-Miner tool-based approach comes into play by offering
automated support for identifying the abstractions of different AORE techniques (e.g.,
viewpoints [20] based, use case [21] based) and helping to build the models. The
tool's automated support helps to reduce the time spent to:
•
Identify model abstractions:
For example concepts such as use cases,
viewpoints, and early aspects that belong to a specific requirements technique
(e.g., Use Case based AORE [22], Viewpoints based AORE [16, 17]) can be
automatically mined from different elicitation documents;
Structure abstractions into various models:
The tool offers features to edit
the identified abstractions (add, remove, filter) as well as to map them into a
chosen model (e.g., a structured AORE specification based on viewpoints or use
cases).
It is important to mention that EA-Miner's automated support does not replace the
work of the requirements engineer but only helps him/her to save time by focusing on
key information. The key insight for early-stage requirements automation is the use of
natural language processing (NLP) to reason about properties of the requirements as
well as the utilization of semantics revealed by the natural language analysis in
building the models. The use of NLP techniques to help with AORE automation was
initially investigated by our previous work [23, 24] and provided some insights (e.g.,
which NLP techniques could be used for identifying model concepts) that helped us to
reach the current state of the tool's implementation. After this, we have added several
features on the tool such as synonym and stemming filtering, frequency analysis, and
support for functional crosscutting as well as made several improvements on the
identification mechanisms. Moreover, we have conducted several case studies
including an industrial case study to evaluate the tool.
Most AORE approaches [16-19, 25] do not provide tool support for the
identification of early aspects from requirements documents with the exception of
Theme/Doc [15]. Therefore, EA-Miner offers a key contribution to complement these
approaches by automating the identification task for them. Moreover, our NLP-based
mining analysis and techniques used (Sects. 3,4) offer a higher degree of automation
when compared to other mining approaches (Sect. 7) as the input requested from the
user is minimal.
The remainder of this paper is structured as follows. Section 2 explains how EA-
Miner can be utilized in an AORE process. Section 3 gives an overview of the
utilization of natural language techniques for requirements model automation. Section
4 shows how EA-Miner uses these NLP techniques to automate the identification of
concepts and mapping of models. Section 5 evaluates the tool showing its time-
effectiveness and also presents data regarding the quality of the produced models.
Moreover, an industrial case study shows how the tool can perform in a real-world
setting and what benefits it can bring for the development process such as identifying
relevant concerns that were missed by domain experts. Section 6 provides further
discussion of EA-Miner and its features. Section 7 presents an overview of existing
related work while Sect. 8 concludes the paper.
•
Search WWH ::
Custom Search