Repository navigation
Correct clinical labeling policy and prepare portable tagger 0.1.2 - #3
Draft
Kkuntal990 wants to merge 4 commits into
Draft
Kkuntal990 wants to merge 4 commits into
Kkuntal990 wants to merge 4 commits into
Conversation
Kkuntal990
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Installed taggers previously lacked bundled resources, failed calls could look like successful abstentions, and clinical rules incorrectly equated MCI with dementia or DBS with Parkinson's disease. Version 0.1.2 combines portable packaging and explicit failure handling with corrected clinical normalization/instructions, one uncalibrated evidence-strength confidence rubric, primary-first label ordering, and machine-generated/unreviewed-at-generation provenance. The 15 examples and allowed label lists remain unchanged; the prompt and MCI alias change prospectively.
Request identities, tested CSV/catalogue writers and noneditable wheel resources support reproducible reuse. The original code uses MIT, with third-party dataset/example rights retained. Consuming applications must display machine origin and support later curator review; this release does not change the public catalogue interface or retag production.
Validation: 120 offline tests pass on Linux/Python 3.11–3.13 in both push and PR CI, with 38% whole-package statement coverage. Normal wheel smoke checks verify bundled corrected policy, machine-origin provenance, all three CLI entry points, successful abstention and malformed-output failure. Regression tests establish software behavior, not improved model accuracy. Historical agreement and repeatability results used the historical prompt and do not validate the corrected one.
Draft pending release publication. No deployment or production writes occurred. Provider/scraping paths are incompletely exercised; HTTP existence checks do not make catalogue upserts atomic. An old embedded database credential was removed earlier in this branch and needs maintainer rotation if still valid.