Title: Requirements' complexity ranking using natural language processing and complexity class correlation with defect severity

Authors: Mukundan Sundararajan; Priti Srikrishnan; Kiran Nayak

Addresses: IBM India Pvt. Ltd., Bengaluru, 560045, India ' IBM India Pvt. Ltd., Bengaluru, 560045, India ' IBM India Pvt. Ltd., Bengaluru, 560045, India

Abstract: This paper addresses the risks to delivery schedule and product quality from non-periodic temporal detection of high severity defects in software projects. The non-periodicity in time and lack of a time boundary in detecting severe defects primarily stems from subjective scheduling of development and testing of product features. One solution is to objectively determine complexity and ranking of requirements to drive the project development and test sequence that uncovers high severity defects early in the life cycle phases. Requirements complexity is strongly correlated with defect severity as measurements show. Applying natural language processing, key words are identified in the given set of requirements, their weights measured to determine the complexity class distribution and ranking that drives the scheduling. The complexity and defect correlation-based sequencing mitigates the risks by discovery of high severity defects in a temporal saw tooth pattern providing the project team sufficient time to fix defects and mitigate the risks.

Keywords: requirements complexity; requirements ranking; requirement complexity classes; natural language processing; NLP; defect severity correlation; defect spatial distribution; defect temporal distribution.

DOI: 10.1504/IJFSE.2020.110591

International Journal of Forensic Software Engineering, 2020 Vol.1 No.2/3, pp.180 - 192

Received: 30 Nov 2018
Accepted: 01 Apr 2019
Published online: 26 Oct 2020 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article