File(s) under permanent embargo
A Comprehensive Requirement Capturing Model Enabling the Automated Formalisation of NL Requirements
journal contribution
posted on 2023-01-30, 03:54 authored by M Osama, A Zaki-Ismail, Mohamed AbdelrazekMohamed Abdelrazek, J Grundy, A IbrahimFormalising natural language (NL) requirements is essential to have formal specifications that enable formal checking and improve the quality of requirements. However, the existing formalisation techniques require engineers to (re)write the system requirements using a set of requirements templates with predefined and limited structure and semantics. The main drawback of using such templates, usually with a fixed format, is the inability to capture diverse requirements outside the scope of the template structure. To address this limitation, a comprehensive reference model is needed to enable capturing key requirement properties regardless of their format, order, or structure. NLP technique can then be used to convert unrestricted NL requirements into the reference model. Using a set of transformation rules, the reference model representing the requirements can be transformed into the target formal notation. In this paper, we introduce requirement capturing model (RCM) to represent NL requirements by adapting to their key properties and without imposing constraints on how the requirements are written. We also implemented a requirements formalisation approach that supports transforming RCM into temporal logic (TL). In addition, we developed an automated similarity checking approach to check the correctness of the constructed RCM structures against the source NL requirements. We carried out extensive evaluation of RCM by comparing it against 15 existing requirements representation approaches on a dataset of 162 requirement sentences. The results show that RCM supports a much wider range of requirements formats compared to any of the existing approaches.
History
Journal
SN Computer ScienceVolume
4Article number
57Pagination
57-Publisher DOI
ISSN
2662-995XeISSN
2661-8907Language
enIssue
1Publisher
Springer Science and Business Media LLCUsage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC