Methodology: Pipeline
The pipeline is orchestrated by a stage dependency graph and runs monthly (on the 1st of each month) via GitHub Actions. The graph currently includes 17 stages, with several optional stages that only run when their inputs or caches are available:
- Extract DBLP data — download and parse the DBLP XML database (~3 GB compressed) for author-paper mappings and affiliations
- Generate statistics — scrape artifact evaluation results from sysartifacts/secartifacts websites and USENIX technical session pages
- Collect repository statistics — gather GitHub stars/forks and archive download counts for artifact repositories
- Check artifact URL availability — run monthly liveness checks over artifact URLs
- Generate AE participation statistics — compare AE participation with DBLP publication counts
- Compute author statistics — match papers to DBLP authors, filter by AE-active years, and compute per-author metrics
- Integrate ArtiFinder-discovered links — ingest automatically discovered artifact links from the ArtiFinder-Data dataset
- Generate area-specific author data — separate systems, security, and combined stats
- Aggregate committee statistics — compile AE membership and chair roles from scraped data
- Compute combined rankings — apply weighted scoring formula, enforce minimum threshold, assign ranks with dense tie-breaking; ArtiFinder-discovered links are reported separately and do not affect scores
- Aggregate institution rankings — sum across affiliated authors and classify institution roles
- Build author profiles — generate per-author detail records
- Build search index — generate the searchable data set and mark automatically discovered ArtiFinder links in search/profile views
- Record ranking history — snapshot current rankings for trend tracking
- Generate visualizations — build SVG charts for the methodology pages and summaries
- Collect paper citations via DOI — fetch citation counts from OpenAlex/Semantic Scholar
- Collect baseline citations — fetch citation counts for non-AE papers for comparison
Independent stages run in parallel where the dependency graph allows. ArtiFinder links are unverified and excluded from all AE-based rankings; they are surfaced separately in search and the dedicated ArtiFinder page. Optional stages may be skipped when their prerequisites are unavailable or their caches are still valid. All output data structures are formally defined in the Data Schemas documentation.
The complete pipeline takes ~30 minutes and processes 3118+ artifact papers.