Current numbers
Live state from the most recent pipeline build baked into
data.js. To refresh after a re-run:
just pipeline && just export && just review-queue && just docs.
Pipeline tables
Row counts in the canonical SQLite, grouped by where the rows came from (raw imports, derived match candidates, final resolutions).
Resolution confidence
How much coverage do you get from curating top-N?
CORDIS participation is heavy-tailed. The chart below plots cumulative
appearance coverage (% of all unresolved appearances) against the number
of top entries you curate, on a log-scale x-axis. Drag the slider to
pick a depth; the readout below says exactly what landing those entries
in pic_ror_overrides.yaml would buy you.
Top of the review queue
The most-mentioned PICs that aren't auto-resolved by the pipeline, sorted by total occurrence across HE, H2020, and FP7. Each row is one curation decision worth making. Click a ROR id to open it on ror.org.
AI-assisted review
Resolution rules used by the pipeline
The resolver tries strategies in priority order and tags the chosen one
on each output row. Counts below are the number of resolutions tagged
with that source in the current build. The strategy names
map to specific match rules; see
organizations-pipeline.md for the
full catalog.