Automated goal model generation from user stories using Large Language Models

In agile software development, user stories capture stakeholder needs but often fail to represent complex requirement relationships. Goal modeling addresses this by linking high-level goals to specific requirements, but manually transforming user stories into goal models is challenging. This research explores using Large Language Models (LLMs) to automate goal model generation through multi-step prompt engineering. LLMs extract intentional elements—goals, tasks, actors, and resources—and generate Goal-oriented Requirements Language (GRL) models compatible with tools like jUCMNav. The study evaluates GPT-4, Llama, and Cohere, focusing on syntactic completeness and correctness. GPT-4 outperforms others, particularly in extracting implicit goals and soft goals, but struggles with intricate relationships like means-end and contribution links. Despite limitations, LLMs show promise in automating labor-intensive aspects of goal modeling, making the process more efficient. This research highlights their potential to support requirements engineers and integrate goal modeling into agile workflows.

Name thesis:degree_name

Master of Applied Science (MASc)

Discipline thesis:degree_discipline

Electrical and Computer Engineering

Grantor

University of Ontario Institute of Technology

Year dc:date.issued

2025

Author dc:creator

• Siddeshwar, Vaishali

Advisors dc:contributor.advisor

• Alwidian, Sanaa
• Makrehchi, Masoud

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10155/1942

OAI identifier oai:identifier

oai:ontariotechu.scholaris.ca:10155/1942