All notable changes to this project will be documented in this file.
- Add Github Action that build & pushes to PYPI index
- Added
needs_review
filter formerge_labels
. - RELEASE_CRITERIA.md
- CHANGELOG.md
- CODE OF CONDUCT.md
- Release automation using Travis
- Added modularized config structure
- New
join_observations
utility - Ability to configure the number of extractors in Makefile
- Feature Lists for sqwiki, kowiki, fiwiki, rowiki, zhwiki and glwiki
extractors
typo in generated parameters- Flaky config test
- Notebook fixes for revscoring 2.x
- Silence mediawiki API errors in
autolabel
. merge_labels
degenerate bug.- Argument order for
merge_label
utility - Minor issue with autolabel mw session args
- Bumped revscoring version to 2.5.1
- Refactored templates to be more modular
- Makefile now reads
trusted_edits
config. - Use
min_samples_leaf
as a tuning param for GB. - Interpret boolean string literals in label merge.
- Make option of auto-labeled dataset optional in
merge_labels
. - Standardized
generate_make
utility and args. - Use informals and badwords for wikidata comment parsing.
- Assume goodfaith in
fetch_labels
. - Exclusions for huwiki badwords/informals
- Added
filter_rate_at_recall
metrics back to Makefile - Migrate from nosetest to pytest
- XGBoost requirement
- Added
feature_lists
for svwiki and cswiki - Added
--filter
param tofetch_labels
utility - Added "about.py" variables (
__version__
,__author__
, etc.)
- Data normalization issues in autolabel utility
- Utilities now use JSON-lines format
- Renamed
label_reverted
toautolabel
- Bumped revscoring requirement to 1.2.8 f9739b7
test_statistics
model param for damaging and goodfaith filtering
- XGBoost from models to tune
- huwiki feature lists for revert model
- ruwiki feature lists for damaging features
- Bumped revscoring requirement to >= 1.2.2
- Updated Revert detection demo notebook with more context.
- Added XGBoost to model tuning
- Added Revert detection demo notebook
- Added Wikibase feature lists
- Added separate Wikitext feature lists
- Update mediawiki features for comment new line regex and user registration.
- Update
wikidatawiki
to use new Wikibase features - Update all damaging feature to use new Wikitext features
- Updated feature lists to handle linear models
- Added edit summary support for
label_reverted
util. New args:exclude-reverted
andexcluded-reverting
- dump reader
- Parallelized
label_reverted
util. - Parallelized
prelabel
util. - enwiki and wikidatawiki make commands use revscoring 1.0.1-rc1
- Remove old linear svc models from Makefile
- configs for models that exclude problematic user features
- informal words from itwiki's damaging feature
- Added
--include-revid
and cut -f2- to all feature extractions in Makefile. - Configuration for model tuning
- Commands for tuning in Makefile
- Import bug for idwiki features in Makefile
- Bumps revscoring requirement to 0.7.10
- Added back user models next to non-user models.
- Resolved
self_reverted
detection bug inlabel_reverted
- added handling for
reverted_to
condition inlabel_reverted
- New
rev_reverted
rule to makefile for all languages. - Added revert labeling to all wikis
- Handles deleted users in
prelabel
utility - Bumped revscoring dep to >= 0.7.2
- Updates makefile to include
rev_ids
in feature extraction
- New
fetch_labels
utility - Added enwiki, fawiki and ptwiki feature lists
- Model References in Makefile
- Minor cleanup in
prelabel
docstring
- Prelabel docstring fixes
- Authors typo
- Module error due to missing
__init__
file.
- First release containing working prelabel script