Toward automated requirements engineering: empirical and architectural foundations for structured parsing and knowledge discovery

Requirements Engineering (RE) relies heavily on natural language, which is often vague, inconsistently structured, and difficult to automate reliably. This thesis presents TRAC-RE, a Traceable, Reliable, Auditable, and Contextual framework for automated requirements engineering. The framework consists of four interconnected phases addressing requirement identification, audit-grade keyword extraction, structured large language model (LLM) parsing, and implicit keyword discovery. A context-aware requirement identification model improved F1 from 0.664 to 0.894 on 110 real-world FinTech and SaaS documents. An audit-grade keyword extraction pipeline achieved perfect precision while revealing a structural coverage ceiling of 1.5 canonical keywords per artifact. A governed High-Level JSON (HLJ) parsing pipeline improved tag precision from 0.657 to 0.897. Finally, a multi-signal implicit discovery engine grounded in a 13,725-entry domain dictionary addressed limitations of explicit extraction. Together, these contributions establish empirical and architectural foundations for reliable, structured, and auditable automated requirements engineering.

Name thesis:degree_name

Master of Applied Science (MASc)

Discipline thesis:degree_discipline

Software Engineering

Grantor

University of Ontario Institute of Technology

Year dc:date.issued

2026

Author dc:creator

• Patel, Dvip

Advisor dc:contributor.advisor

• Alwidian, Sanaa

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10155/2109

OAI identifier oai:identifier

oai:ontariotechu.scholaris.ca:10155/2109