PhD Defense: 'Requirements classification and multi-criteria prioritization using AI techniques'

This thesis focuses on improving two key activities in Requirements Engineering: requirements classification and prioritization. These activities are addressed within the context of Software Product Lines, which add complexity due to their variability and scale. The work tackles challenges such as managing large volumes of requirements and relying on a limited number of domain experts, particularly when requirements are written in Spanish, a language with limited annotated datasets and NLP resources.

Our research proposes and evaluates machine learning and natural language processing (NLP) techniques to automate these activities, enhancing efficiency and reducing stakeholder dependency. Using datasets such as PROMISE and ReSpa, the results demonstrate the feasibility of automating these processes and highlight ensemble approaches, pre-trained language models (e.g., BETO), and learning-to-rank algorithms (e.g., LambdaMART) as effective solutions for Spanish-language software product line environments.

Supervisors: Nelly Condori Fernández and Miguel Luaces