Investigating Metadata Practices for Repository and Donation Platform URLs in PyPI Libraries: An Empirical Study

<h2 dir="ltr">Overview</h2><p dir="ltr">Replication package for a full research paper enabling end-to-end reproduction of:</p><ol><li>LLM-based topic modeling pipeline of open-ended survey responses.</li><li>Two quantitative results replicated from prior work using curated Jupyter notebooks.</li></ol><h2 dir="ltr">Contents</h2><ul><li><b>Surveys:</b> PDF copies of both survey instruments.</li><li><b>Evaluation Forms:</b> HTML forms for human rating of topic quality and associated e-mail communications for recruiting human raters, distributing the forms, and acknowledging participation.</li><li><b>Replication Package:</b> Two notebooks to replicate quantitative results (GitHub repository accessibility for PyPI libraries; Donation platform usage for PyPI libraries) with dataset, requirements and instructions for execution.</li><li><b>Topic-Modeling Pipeline:</b> Kedro project with configs for repository and donation platform survey answers, and LLM integration via Ollama.</li></ul><h2 dir="ltr">Reproduction</h2><ul><li>Step-by-step instructions and system setup are documented in the top-level README.</li><li>Components can be reproduced independently (pipeline or notebooks).</li></ul><h2 dir="ltr">Key Outputs</h2><ul><li>HTML reports, metrics, and JSON summaries under the project’s reporting folders (e.g., data/08_reporting/…).</li><li>Packaged evaluation HTML for convenient human rating.</li></ul><h2 dir="ltr">System Requirements</h2><ul><li>Python 3.10; Kedro CLI.</li><li>Jupyter (for notebooks).</li><li>LLM: Ollama and selected model.</li><li>Tested hardware: 2x NVIDIA L40S, AMD EPYC 9354P 32-Core Processor, 188GB RAM</li></ul><p><br></p>