A onn-line tool to help teams that use agile practices to building better user stories.
The User Story Tutor (UST) takes a User Story text in English as input and provides personalized recommendations for improvement, supported by a large language model (LLM). Additionally, the tool offers a Story Points prediction generated by a machine learning algorithm trained on data from other projects. The user also receives readability indexes for the User Story, which can serve as indicators of text clarity.
UST supports agile teams in the creation of User Stories and assists in the development process during both the preparation phase and task effort estimation. UST is a web application accessible via browser on mobile devices, PCs, and notebooks. It utilizes an LLM provided by OpenAI to recommend improvements and present readability indexes.
The application is organized into three modules, each with well-defined functions. The Recommender module handles User Story recommendation requests. It manages the prompt, combines it with the new User Story text, queries OpenAI via API, and prepares the response for presentation to the user. The module responsible for estimating Story Points employs a predictive model trained on historical data to assist developers with their estimates, providing a reference for the team in effort estimation. Readability indexes are extracted from the text using basic natural language processing techniques.
This dataset includes User Story (or Issue) text descriptions, User Story titles, and Story Points from 33 software development projects, comprising a total of 20,479 User Stories (or issues) extracted from GitLab repositories, amounting to 12,262.7 Story Points. The mining process focused on GitLab’s top open-source projects that use agile software development methodologies and record task sizes in Story Points. Only tasks with the State attribute set to Closed and with the Weight attribute filled in were collected. The Weight field in GitLab is used to record the effort in Story Points. The data was mined between January 2023 and April 2023. The projects in the dataset have diverse characteristics, covering different programming languages, business domains, and geographic locations of the teams.
The paper below shortly this dataset in section 4.1 Dataset
Neo, G. S.; Moura, J.; Almeida, H.; Neo, A. and Freitas Júnior, O. (2024). User Story Tutor (UST) to Support Agile Software Developers. In Proceedings of the 16th International Conference on Computer Supported Education - Volume 2: CSEDU; ISBN 978-989-758-697-2; ISSN 2184-5026, SciTePress, pages 51-62. DOI: 10.5220/0012619200003693
For coding, we used StreamLit, a library for building open-source applications for machine learning and data science. The Recommender Module performs a query to OpenAI. The scikit-learn libraries were also used.