Repository navigation
Import pilot-1 annotations with label origin; first pilot dev scores - #20
Merged
Merged
Conversation
…v scores Annotation records gain an `origin` (manual, model_draft_accepted, model_draft_edited) and `origin_model`, read from Label Studio's parent_prediction link; only label changes count as edits, not notes. Schema version 1.1 (additive). The import report counts origins, and `evaluate --gold-origin manual` scores hand-made records only. pilot-1: 100 clean records (22 manual, 78 accepted model drafts). Rules and CRF are scored on the dev split only; the locked test split is untouched. Reports separate the manual slice from the all slice, which measures agreement with the draft model rather than accuracy.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
AdvisoryAnnotation.origin(manual/model_draft_accepted/model_draft_edited) andorigin_model, read from Label Studio'sparent_predictionlink. Only label changes (spans, categories, roles, relations, document status) count as edits; the notes field does not.--gold-origin manualscores hand-made records only; the report header states the slice and document count.annotations/pilot/pilot-1.jsonl: 100 clean records, 0 issues (22manual, 78model_draft_acceptedfrommodel-pilot-draft-2026-11-01-v1).reports/pilot/: rules-2 and crf-1 on the dev split only,manualandallslices. Locked test split untouched.annotations/README.mdupdated: the label-origin principle, pilot status, and a preliminary dev-only results table.Preliminary dev results (
manualslice, 10 docs, 17 gold conditions)Checks
ruff, ruff format, mypy strict,
pytest -m "not slow"(212 passed). CRF retrain reproduces predictions byte for byte.