Automated Features Extraction from Software Requirements Specification (SRS) Documents as The Basis of Software Product Line (SPL) Engineering

Abstract

Extractive Software Product Line Engineering (SPLE) puts features on the foremost aspect in domain analysis that needs to be extracted from the existing system's artifact. Feature in SPLE, which is closely related to system functionality, has been previously studied to be extracted from source code, models, and various text documents that exist along the software development process. Source code, with its concise and normative standard, has become the most focus target for feature extraction source on many kinds of research. However, in the software engineering principle, the Software Requirements Specification (SRS) document is the basis or main reference for system functionality conformance. Meanwhile, previous researches of feature extraction from text document are conducted on a list of functional requirement sentences that have been previously prepared, not literally SRS as a whole document. So, this research proposes direct processing on the SRS document that uses requirement boilerplates for requirement sentence statement. The proposed method uses Natural Language Processing (NLP) approach on the SRS document. Sequence Part-of-Speech (POS) tagging technique is used for automatic requirement sentence identification and extraction. The features are acquired afterward from extracted requirement sentences automatically using the word dependency parsing technique. Besides, mostly the previous researches about feature extraction were using non-public available SRS document that remains classified or not accessible, so this work uses selected SRS from publicly available SRS dataset to add reproducible research value. This research proves that requirement sentence extraction directly from the SRS document is viable with precision value from 64% to 100% and recall value from 64% to 89%. While features extraction from extracted requirement sentences has success rate from 65% to 88%.

The Use of NLP-Based Text Representation Techniques to Support Requirement Engineering Tasks: A Systematic Mapping Review

“…• Extracting Requirements from Textual Documents: 7 papers [92,93,94,95,96,97,98] presented approaches to extract requirements from SRS documents, policies, user manuals, and emails.…”

“…(2) Representing each requirement as a pre-defined set of linguistic features. (3) Then, proceeding to the solving phase which uses ML techniques (such as Decision Trees [97,138,81,118,52], SVM [53,10,88], and RF [96,60,90,49]), or rule-based approach using a set of syntactic regular expressions [125,95,126,130,117]. The used pre-defined set of features (in step 2) usually includes features related to the following four groups:…”

“…-Extracting requirements from documents: the structured format of these documents motivated many researchers to use lexical and syntactic features. All retrieved papers under this sub-category used lexical and syntactic representations [92,96,95].…”

A BERT-based transfer learning approach to text classification on software requirements specifications

“…Substantial amount of research have been dedicated to requirement analysis such as classifying and prioritizing requirements tasks using NLP [2][3][4][5][6]. However, most automated solutions made use of information retrieval and supervised machine learning techniques for the NLP tasks [6].…”

A Semantic Comparison of Feature Requirements Extraction Methods

“…In previous research, Haris conducted research on automatic requirement extraction on the Software Specification Document (SRS) Document as the basis of the software product line [5] . This study focuses on using NLP to extract sentences based on boilerplate templates with the Part-of-Speech (POS) Tagging method, which will later become the basis of the feature.…”