Repository logo
 

Humanizing the Release Notes Generation Process

Date

2025-03-05

Journal Title

Journal ISSN

Volume Title

Publisher

ORCID

0000-0003-1509-6327

Type

Thesis

Degree Level

Doctoral

Abstract

Software release notes play a crucial role in software development by providing users and stakeholders with essential information about new features, bug fixes, performance improvements, and security updates. However, despite their importance, release notes often suffer from various documentation issues, including inconsistency, redundancy, lack of clarity, and missing traceability links. These problems can negatively impact software usability, maintenance, and user experience, leading to confusion among developers and end-users. The manual effort required to produce high-quality release notes further exacerbates these challenges, especially in large and fast-paced software development environments. This thesis addresses these challenges by leveraging natural language processing (NLP) and machine learning (ML) techniques to automate the release note generation process, improve documentation quality, and enhance traceability. Specifically, we investigate: (1) the detection and mitigation of documentation anti-patterns by using multi-label classification models that hinder the readability and usefulness of release notes, (2) automated traceability link recovery between release notes and commit messages, issues, and pull requests to improve documentation completeness by considering what, why and how information, and (3) a content-tailoring approach that generates structured and concise release notes while preserving critical information. To achieve these objectives, we fine-tuned state-of-the-art transformer-based language models, e.g., BART, on domain-specific software repositories using commit messages, pull request and issue titles. We also develop PytextQltEval, a novel framework for evaluating text quality attributes such as readability, conciseness, consistency, and structuredness. This framework incorporates both automated metrics and human-centered evaluations to assess the effectiveness of the generated release notes. As part of our future work, we aim to integrate our approach into real-world software development workflows by developing a GitHub-based tool and publishing it in the GitHub Marketplace. This tool will allow open-source contributors and software teams to automate the generation of high-quality release notes while maintaining adaptability through customization configurations.

Description

Keywords

Release note, NLP, human perspectives

Citation

Degree

Doctor of Philosophy (Ph.D.)

Department

Computer Science

Program

Computer Science

Advisor

Committee

McCalla, Gord;Vassileva, Julita;Khan, Shahedul;Msintosh, Shane;McQuillan, Ian

Part Of

item.page.relation.ispartofseries

DOI

item.page.identifier.pmid

item.page.identifier.pmcid