Requirements-Driven Software Quality Assurance using Large Language Models (LLMs)

Software Quality Assurance (SQA) is a critical aspect of software development, ensuring that the final product meets specified requirements and performs reliably. However, most SQA activities, even in practice, are often decoupled from the Requirements Engineering (RE) stage, leading to gaps in the testing process, quality issues and, eventually, major software project failures. This PhD research aims to bridge this gap by leveraging Large Language Models (LLMs) to generate test artefacts directly from requirements. By focusing on requirements-driven SQA, this PhD project seeks to enhance the alignment between software requirements and testing processes.

The research will investigate using LLMs to automatically generate test scenarios and other test artefacts from natural language (NL) requirements, e.g., user stories. The approach involves fine-tuning LLMs to understand and interpret various requirements, translating them into comprehensive test artefacts that can be used throughout the SQA process. This methodology aims to improve the accuracy and efficiency of test artefact generation, ensuring that all critical aspects of the requirements are thoroughly tested.

The expected outcomes of this research include a novel framework for integrating LLMs into the SQA process, enhanced methodologies for generating test artefacts from requirements, and empirical validation of the proposed approach through case studies and experiments. By addressing the decoupling issue in traditional SQA practices, this research has the potential to significantly improve software quality, reduce development costs, and increase the overall reliability of software systems.

Project Lead

Fanyu Wang (PhD candidate)

Supervisors

Dr Chetan Arora, Dr Chakkrit Tantithamthavorn, Prof Aldeida Aleti