Correction front — where editorial attention moves over time
The diachronic correction front: month-by-month and era-split views of where
Cologne dictionaries receive accepted fixes, coloured by OBS-T microstructure
component (headword, sense, markup, citation, …). Atlas renders this
strip; the event typology is owned by
csl-observatory
(OBS-T / MW-ATTENTION boundary — never recomputed here).
Sister pages: spatial loci on Correction loci, shared-error lineage overlay on Lineage Sankey.
Trust Block
- Evidence: OBS-T
correction_events_release.csvaggregated bynpm run build-correction-lane-overlaysintosrc/data/corrections/correction_front.json(events; top dicts in the strip). - Limitations: event grain ≠ csl-corrections change-file loci;
error_component=unattributedis the plurality class; corrector identity omitted (personal data). - Validation:
npm run build-correction-lane-overlays+npm test(correction-lane-overlays); page bynpm run build. - Owner repo:
csl-atlas(rendering). Data owner:csl-observatory. - Next use: teach the 2014–2018 vs 2019–2026 component shift; pair with correction-loci for "where in the book" vs "when / which layer".
Era overview — 2014–2018 vs 2019–2026
Total events per top dictionary in each era (all components). The second era is not simply "more of the same": component mix shifts (see next chart).
Era × component (small multiples)
Stacked component share per dictionary within each era. Read left→right for the front's maturation: surface/markup vs sense-layer attention.
Monthly front — pick a dictionary
Month × component stream for one top dictionary. Spikes are usually batch ingest or focused proofreading campaigns, not a smooth editorial rate.