Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,9 @@ ipython/.ipynb_checkpoints/
# Temporary text editor files
*~
.*.sw?
.ipynb_checkpoints/
.vscode/
*.ipynb

# C extensions
*.so
Expand Down
32 changes: 17 additions & 15 deletions Makefile
Original file line number Diff line number Diff line change
Expand Up @@ -46,7 +46,7 @@ models: \
srwiki_models \
svwiki_models \
tawiki_models \
translatewiki_models \
# translatewiki_models \
trwiki_models \
ukwiki_models \
urwiki_models \
Expand Down Expand Up @@ -96,7 +96,7 @@ tuning_reports: \
srwiki_tuning_reports \
svwiki_tuning_reports \
tawiki_tuning_reports \
translatewiki_tuning_reports \
# translatewiki_tuning_reports \
trwiki_tuning_reports \
ukwiki_tuning_reports \
urwiki_tuning_reports \
Expand Down Expand Up @@ -2448,7 +2448,7 @@ models/jawiki.goodfaith.gradient_boosting.model: \
-p 'learning_rate=0.1' \
-p 'max_depth=1' \
-p 'max_features="log2"' \
-p 'min_samples_leaf=7' \
-p 'min_samples_leaf=3' \
-p 'n_estimators=300' \
--label-weight $(goodfaith_weight) \
--pop-rate "true=0.00461076560154407" \
Expand Down Expand Up @@ -2522,10 +2522,11 @@ models/kowiki.damaging.gradient_boosting.model: \
editquality.feature_lists.kowiki.damaging \
damaging \
--version=$(damaging_major_minor).0 \
-p 'learning_rate=0.01' \
-p 'max_depth=7' \
-p 'learning_rate=0.5' \
-p 'max_depth=1' \
-p 'max_features="log2"' \
-p 'n_estimators=700' \
-p 'min_samples_leaf=5' \
-p 'n_estimators=500' \
--label-weight $(damaging_weight) \
--pop-rate "true=0.038665452792802445" \
--pop-rate "false=0.9613345472071976" \
Expand Down Expand Up @@ -2559,7 +2560,8 @@ models/kowiki.goodfaith.gradient_boosting.model: \
-p 'learning_rate=0.01' \
-p 'max_depth=7' \
-p 'max_features="log2"' \
-p 'n_estimators=700' \
-p 'min_samples_leaf=13' \
-p 'n_estimators=500' \
--label-weight $(goodfaith_weight) \
--pop-rate "true=0.9807743801210304" \
--pop-rate "false=0.019225619878969646" \
Expand Down Expand Up @@ -4063,11 +4065,11 @@ models/zhwiki.damaging.gradient_boosting.model: \
editquality.feature_lists.zhwiki.damaging \
damaging \
--version=$(damaging_major_minor).0 \
-p 'learning_rate=0.01' \
-p 'max_depth=3' \
-p 'learning_rate=0.1' \
-p 'max_depth=1' \
-p 'max_features="log2"' \
-p 'min_samples_leaf=7' \
-p 'n_estimators=700' \
-p 'min_samples_leaf=3' \
-p 'n_estimators=300' \
--label-weight $(damaging_weight) \
--pop-rate "true=0.0405" \
--pop-rate "false=0.9595" \
Expand Down Expand Up @@ -4098,11 +4100,11 @@ models/zhwiki.goodfaith.gradient_boosting.model: \
editquality.feature_lists.zhwiki.goodfaith \
goodfaith \
--version=$(goodfaith_major_minor).0 \
-p 'learning_rate=0.01' \
-p 'max_depth=3' \
-p 'learning_rate=0.1' \
-p 'max_depth=1' \
-p 'max_features="log2"' \
-p 'min_samples_leaf=5' \
-p 'n_estimators=500' \
-p 'min_samples_leaf=3' \
-p 'n_estimators=300' \
--label-weight $(goodfaith_weight) \
--pop-rate "true=0.9682" \
--pop-rate "false=0.03180000000000005" \
Expand Down
2 changes: 1 addition & 1 deletion config/wikis/jawiki.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -62,5 +62,5 @@ models:
max_depth: 1
max_features: log2
n_estimators: 300
min_samples_leaf: 7
min_samples_leaf: 3

10 changes: 6 additions & 4 deletions config/wikis/kowiki.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -45,10 +45,11 @@ models:
cv_train:
algorithm: GradientBoosting
parameters:
learning_rate: 0.01
max_depth: 7
learning_rate: 0.5
max_depth: 1
max_features: log2
n_estimators: 700
n_estimators: 500
min_samples_leaf: 5
goodfaith:
observations: labeled_revisions.w_cache.20k_2016
label: goodfaith
Expand All @@ -60,4 +61,5 @@ models:
learning_rate: 0.01
max_depth: 7
max_features: log2
n_estimators: 700
n_estimators: 500
min_samples_leaf: 13
16 changes: 8 additions & 8 deletions config/wikis/zhwiki.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -43,11 +43,11 @@ models:
cv_train:
algorithm: GradientBoosting
parameters:
learning_rate: 0.01
max_depth: 3
learning_rate: 0.1
max_depth: 1
max_features: log2
n_estimators: 700
min_samples_leaf: 7
n_estimators: 300
min_samples_leaf: 3
goodfaith:
observations: labeled_revisions.w_cache.100k_2016
label: goodfaith
Expand All @@ -56,8 +56,8 @@ models:
cv_train:
algorithm: GradientBoosting
parameters:
learning_rate: 0.01
max_depth: 3
learning_rate: 0.1
max_depth: 1
max_features: log2
n_estimators: 500
min_samples_leaf: 5
n_estimators: 300
min_samples_leaf: 3
2 changes: 1 addition & 1 deletion editquality/about.py
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
__name__ = "editquality"
__version__ = "0.5.0"
__version__ = "0.5.1"
__author__ = "Aaron Halfaker"
__author_email__ = "ahalfaker@wikimedia.org"
__description__ = "A library for performing automatic detection of the " + \
Expand Down
2 changes: 1 addition & 1 deletion editquality/feature_lists/jawiki.py
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@
]

damaging = wikipedia.page + \
wikitext.parent + wikitext.diff + mediawiki.user_rights + \
wikitext.parent_cjk + wikitext.diff_cjk + mediawiki.user_rights + \
mediawiki.protected_user + mediawiki.comment + \
badwords + informals + \
enwiki.badwords + enwiki.informals
Expand Down
4 changes: 2 additions & 2 deletions editquality/feature_lists/kowiki.py
Original file line number Diff line number Diff line change
Expand Up @@ -21,10 +21,10 @@
]

damaging = wikipedia.page + \
wikitext.parent + wikitext.diff + mediawiki.user_rights + \
wikitext.parent_cjk + wikitext.diff_cjk + mediawiki.user_rights + \
mediawiki.protected_user + mediawiki.comment + \
badwords + informals + \
enwiki.badwords + enwiki.informals + enwiki.dict_words
enwiki.badwords + enwiki.informals
"Damaging Features"

reverted = damaging
Expand Down
65 changes: 65 additions & 0 deletions editquality/feature_lists/tests/test_cjk_wiki.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
from revscoring.dependencies import solve
from revscoring.datasources import revision_oriented
from .. import zhwiki
from .. import jawiki
from .. import kowiki


r_text = revision_oriented.revision.text
p_text = revision_oriented.revision.parent.text


def test_zhwiki():
p_text_text = """
中國是位於亞歐大陸的國家或地理區域。此名稱最早见于西周,
回頭見 拜拜 回見
"""
r_text_text = """
中國中國是位於亞歐大陸的國家或地理區域。此名稱最早见于西周,
回頭見 拜拜 回見
"""
cache = {p_text: p_text_text,
r_text: r_text_text}

assert (list(solve(zhwiki.wikitext.diff_cjk, cache=cache)) ==
[2, 1.0, 1.0])
assert ([round(i) for i in solve(zhwiki.wikitext
.parent_cjk, cache=cache)] ==
[4, 3, 2, 1])


def test_jawiki():
p_text_text = """
敗戦後は桑原武夫の『第二芸術-現代俳句について』
(1946年)によって、短詩型である俳句の限界が指摘された。
"""
r_text_text = """
敗戦後は桑原武夫の『第二芸術-現代俳句について』
(1946年)によって、短詩型である俳句の限界が指摘された た。
"""
cache = {p_text: p_text_text,
r_text: r_text_text}
assert (list(solve(jawiki.wikitext.diff_cjk, cache=cache)) ==
[1, 2.0, 1.0])
assert ([round(i) for i in solve(jawiki.wikitext
.parent_cjk, cache=cache)] ==
[4, 3, 1, 1])


def test_kowiki():
p_text_text = """
국어사 자료에서 ‘김치’가 소급하는 최초의 형태는
16세기의 ‘딤치’이다.
"""
r_text_text = """
국어사 자료에서 ‘김치’가 소급하는 최초의 형태는
16세기의 ‘딤치’이다 다.
"""
cache = {p_text: p_text_text,
r_text: r_text_text}

assert (list(solve(kowiki.wikitext.diff_cjk, cache=cache)) ==
[1, 2.0, 1.0])
assert ([round(i) for i in solve(kowiki.wikitext
.parent_cjk, cache=cache)] ==
[3, 3, 2, 1])
72 changes: 72 additions & 0 deletions editquality/feature_lists/wikitext.py
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,25 @@ def _process_new_longest(p_longest, r_longest):
name="revision.parent.markups_per_token"),
]

parent_cjk = [
log(wikitext.revision.parent.cjk_chars + 1),
log(wikitext.revision.parent.cjk.cjks + 1),
log(wikitext.revision.parent.headings + 1),
log(wikitext.revision.parent.wikilinks + 1),
log(wikitext.revision.parent.external_links + 1),
log(wikitext.revision.parent.templates + 1),
log(wikitext.revision.parent.ref_tags + 1),
div(wikitext.revision.parent.cjk_chars,
max(wikitext.revision.parent.cjk.cjks, 1),
name="revision.parent.cjk.cjkchars_per_cjkwordthing"),
div(wikitext.revision.parent.cjk.cjks,
max(wikitext.revision.parent.cjk.tokens, 1),
name="revision.parent.cjk.cjkwordthings_per_token"),
div(wikitext.revision.parent.markups,
max(wikitext.revision.parent.cjk.tokens, 1),
name="revision.parent.markups_per_token"),
]

diff = [
wikitext.revision.diff.markup_delta_sum,
wikitext.revision.diff.markup_delta_increase,
Expand Down Expand Up @@ -91,3 +110,56 @@ def _process_new_longest(p_longest, r_longest):
depends_on=[wikitext.revision.parent.longest_repeated_char,
wikitext.revision.longest_repeated_char])
]

diff_cjk = [
wikitext.revision.diff.markup_delta_sum,
wikitext.revision.diff.markup_delta_increase,
wikitext.revision.diff.markup_delta_decrease,
wikitext.revision.diff.markup_prop_delta_sum,
wikitext.revision.diff.markup_prop_delta_increase,
wikitext.revision.diff.markup_prop_delta_decrease,
wikitext.revision.diff.number_delta_sum,
wikitext.revision.diff.number_delta_increase,
wikitext.revision.diff.number_delta_decrease,
wikitext.revision.diff.number_prop_delta_sum,
wikitext.revision.diff.number_prop_delta_increase,
wikitext.revision.diff.number_prop_delta_decrease,
sub(wikitext.revision.cjk_chars,
wikitext.revision.parent.cjk_chars,
name="revision.diff.cjkchars_change"),
sub(wikitext.revision.cjk.tokens,
wikitext.revision.parent.cjk.tokens,
name="revision.diff.cjk.tokens_change_inc_cjktokens"),
sub(wikitext.revision.cjk.cjks,
wikitext.revision.parent.cjk.cjks,
name="revision.diff.cjk.cjkwordthings_change"),
sub(wikitext.revision.markups,
wikitext.revision.parent.markups,
name="revision.diff.markups_change"),
sub(wikitext.revision.headings,
wikitext.revision.parent.headings,
name="revision.diff.headings_change"),
sub(wikitext.revision.external_links,
wikitext.revision.parent.external_links,
name="revision.diff.external_links_change"),
sub(wikitext.revision.wikilinks,
wikitext.revision.parent.wikilinks,
name="revision.diff.wikilinks_change"),
sub(wikitext.revision.templates,
wikitext.revision.parent.templates,
name="revision.diff.templates_change"),
sub(wikitext.revision.tags,
wikitext.revision.parent.tags,
name="revision.diff.tags_change"),
sub(wikitext.revision.ref_tags,
wikitext.revision.parent.ref_tags,
name="revision.diff.ref_tags_change"),
Feature("revision.diff.longest_new_token",
_process_new_longest, returns=int,
depends_on=[wikitext.revision.parent.longest_token,
wikitext.revision.longest_token]),
Feature("revision.diff.longest_new_repeated_char",
_process_new_longest, returns=int,
depends_on=[wikitext.revision.parent.longest_repeated_char,
wikitext.revision.longest_repeated_char])
]
4 changes: 2 additions & 2 deletions editquality/feature_lists/zhwiki.py
Original file line number Diff line number Diff line change
@@ -1,5 +1,4 @@
from revscoring.languages import chinese

from . import enwiki, mediawiki, wikipedia, wikitext

badwords = [
Expand Down Expand Up @@ -28,8 +27,9 @@
chinese.words_to_watch.revision.diff.match_prop_delta_increase,
chinese.words_to_watch.revision.diff.match_prop_delta_decrease
]

damaging = wikipedia.page + \
wikitext.parent + wikitext.diff + mediawiki.user_rights + \
wikitext.parent_cjk + wikitext.diff_cjk + mediawiki.user_rights + \
mediawiki.protected_user + mediawiki.comment + \
badwords + informals + words_to_watch + \
enwiki.badwords + enwiki.informals
Expand Down
Loading