Recipe for Discovery: Building the UC Open Source Repository Browser From Scratch
summary
Built a relational database from GitHub REST API queries and campus keyword combinations, collecting metadata, READMEs, contributor info and community files; 52,000 repositories found across UC campuses but less than 50% affiliated after manual labels, embeddings + SVMs, and LLM (GPT‑4o) classification; only ~4,800 organizations own repos and PR templates/contributing guides/security policies appear in under 1% of affiliated projects. A manually curated showcase website with search, filters and contact/funder fields surfaces actively maintained, university‑domain projects for outreach and funding. 52,000 repositories found GitHub REST API
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:04Purpose and goals
- 2:27Building the database
- 8:23Repository affiliation
- 10:12Testing affiliation
- 13:29ML and LLM approaches
- 15:38Affiliation analysis results
- 17:04Open source practices
- 20:36UCRB showcase site
- 24:30Demo of features
- 26:44Future plans
- 29:10Resources and involvement
- 29:53Q&A
related talks
14mGitHub's Data Story: Open Source at the Home for Open Source
40mDiscovering Emerging Open Source Communities Through...
31mTelling a Great Story on GitHub
20mFind better cores with LibreCores
45mOSPOs in Academia: Increasing the Impact of Educational Ins...
28mGitHub Gone Wrong - Lessons Learned from Organic Open Source
47mPolicing Open-Source Projects at Scale
13mOpening Address