Hoi,
I noticed the notion about "quality in Wikidata". The approach is very much
in line with what is the norm in Wikipedia. This is inot the right approach
for Wikidata. Many of the items in Wikidata can be of high "quality"; ie
the statements have a source and there are enough labels but the true value
of these items are in the use of these items as statements in other items..
(for instance a university indicates that someone studied there). Another
quality point is that for authors a VIAF statements allows for the linking
in Wikipedias in external sources. This is of a high importance, it makes
Wikidata useful and, if that is not of a quality consideration what is?
One other aspect of Wikidata is that it is still highly immature. Just
consider the statistics for labels and statements [1] . This is only the
first month where less than 10% of our items have no statement.. We talk
about quality but quality should have a practical meaning. Just saying this
or that item is so good, it makes for stamp collecting. The point of a
stamp is not to collect them it is to send mail. Quality means that we know
how many articles have been written in one or more editathons. It is in
finding for a collection of items that it is better known what award,
schooling has been achieved by the people that was written for. It is in
using Wikidata to indicate what categories could be in what Wikipedia
article.
Quality needs to be actionable. What is the use of static quality?
Thanks,
GerardM
[1]
Sharing some good news, both about the progress of
ORES and (my primary
inspiration for sharing this email) significant improvements in article
quality thanks to WikiProject Women scientists. The latter has been
designated as the Keilana Effect.
Pine
---------- Forwarded message ----------
From: Aaron Halfaker <aaron.halfaker(a)gmail.com>
Date: Thu, Mar 16, 2017 at 2:14 PM
Subject: Re: [Wikitech-l] The Revision Scoring weekly update
To: Application of Artificial Intelligence and other advanced computing
strategies to Wikimedia Projects <ai(a)lists.wikimedia.org>
Cc: wikitech-l <wikitech-l(a)lists.wikimedia.org>
Hey folks!
I should really stop calling this a weekly update because it's getting a
bit silly at this point. :) But if it were a weekly update, it would
cover the weeks of 42 - 46.
*Highlights:*
- 3 new models: Finnish Wikipedia (reverted) and Estonian Wikipedia
(damaging & goodfaith)
- We estimated and agreed on funding for ORES servers in the next year
with Operations
- We published a paper about vandalism detection in Wikidata and a blog
post about the massive effect of some initiatives on coverage of Women
Scientists in Wikipedia.
*New development:*
- We added recall-based threshold metrics to the new draftquality model
which should help tool devs know what which new page creations to
highlight
for review[1]
- We added optional notices for ORES pages which will help us visually
distinguish our experimental install in WMFlabs from the Prod install (
ores.wikimedia.org)[2]
- We added basic language support for Finish (Thanks 4shadoww)[3] and
deployed a 'reverted' model[4]
- We lead a discussion in Wikidata about "item quality" that resulted in
a Wikipedia 1.0 like scale for Wikidata quality[5,6] and designed a
Wikilabels form to capture the gist of it[7]
- We enabled the ORES Review Tool on Czech Wikipedia[8]
- We configured ChangeProp to use our new minified JSON output to save
bandwidth[9]
- We extended the Estonian language assets (Thanks Cumbril)[10] and
deployed the 'damaging' and 'goodfaith' models[11,12]
- We enabled a testing model for 'goodfaith' on the Beta Cluster to make
it easier for the Collaboration team to run tests with their new filter
interface[13]
- We created a new "precache" endpoint that will allow us to
de-duplicate configuration with ChangeProp and handle all routing in
ORES
locally[14]
*Resourcing:*
- We completed a 2 year estimate of ORES resource needs and discussed
funding (capital expendature) for ORES in the coming fiscal year[15].
This
will allow us to continue to grow ORES both in number of models and in
scoring capacity.
*Communications:*
- Amir improved the KDD paper based on review feedback[16] and got it
published[17]
- We published a blob post about our measurements of WikiProject Women
Scientists[18,19] -- "The Keilana Effect"
- Thanks to Cumbril's work, the Estonian labeling campaing was
finished[20]
*Deployments:*
- In early February, we deployed a new set of translations to Wikilabels
(specifcally targeting Romanian Wikipedia)[21]
- In mid-February, we deployed some fixes to ORES documentation and
response formatting[22]
- In mid-March, we deployed 3 new scoring models and ORES notices[23]
*Maintenance and robustness:*
- We fixed a serious issue in the "mwoauth" library that Wikilabels
depends on[24]
- We reduced the number of revisions per request that we could receive
via api.php[25]
- We investigated a scap issue that broke ORES deployment[26]
- We fixed a minor issue with JSON minification behavior[27] and
hard-coding of the location of ORES in the documentation[28]
- We improved performance of ORES filters on MediaWiki[29]
- We improved the language describing ORES behavior on
Special:Contributions[30]
- We added a notice to the Wikipages that Dexbot maintains about its
behavior[31]
- We added notices to
ores.wmflabs.org about it's experimental
nature[32]
- We fixed some issues with testing Finnish language assets[33]
- We fixed some styling issues that resulted from an upgrade of OOJS
UI[34]
1.
https://phabricator.wikimedia.org/T157454 -- Add recall based
thresholds
to draftquality model
2.
https://phabricator.wikimedia.org/T150962 -- Add an optional notice to
ORES main and ui pages
3.
https://phabricator.wikimedia.org/T158587 -- Add language support for
Finnish
4.
https://phabricator.wikimedia.org/T160228 -- Train/test reverted model
for fiwiki
5.
https://phabricator.wikimedia.org/T157489 -- [Discuss] item quality in
Wikidata
6.
https://www.wikidata.org/wiki/Wikidata:Item_quality
7.
https://phabricator.wikimedia.org/T155828 -- Design item_quality form
for Wikidata
8.
https://phabricator.wikimedia.org/T151611 -- Enable ORES Review Tool on
Czech Wikipedia
9.
https://phabricator.wikimedia.org/T157693 -- Use minified JSON format
in
ChangeProp
10.
https://phabricator.wikimedia.org/T160193 -- Extend estonian language
assets from Wiki page
11.
https://phabricator.wikimedia.org/T159608 -- Train/test
damaging/goodfaith models for etwiki
12.
https://phabricator.wikimedia.org/T130280 -- Deploy edit quality
models
for etwiki
13.
https://phabricator.wikimedia.org/T160467 -- Enable 'goodfaith' on
testwiki on Beta Cluster
14.
https://phabricator.wikimedia.org/T148714 -- Create generalized
"precache" endpoint for ORES
15.
https://phabricator.wikimedia.org/T157222 -- Estimate ORES capex for
FY2017-18
16.
https://phabricator.wikimedia.org/T148443 -- Improve the KDD paper
based on the review
17.
https://arxiv.org/abs/1703.03861
18.
https://phabricator.wikimedia.org/T160078 -- Blog post about wp10
measurements of Women Scientists
19.
https://blog.wikimedia.org/2017/03/07/the-keilana-effect/
20.
https://phabricator.wikimedia.org/T129702 -- Complete etwiki edit
quality campaign
21.
https://phabricator.wikimedia.org/T157580 -- Deploy Romanian
translations for Wiki labels
22.
https://phabricator.wikimedia.org/T157842 -- Prod deployment of ORES
23.
https://phabricator.wikimedia.org/T160279 -- Deploy ores in prod
(Mid-March)
24.
https://phabricator.wikimedia.org/T157858 -- mwoauth is broken
25.
https://phabricator.wikimedia.org/T157983 -- Reduce the number of
revisions that can be requested in one batch
26.
https://phabricator.wikimedia.org/T157623 -- Investigate failed ORES
deployment
27.
https://phabricator.wikimedia.org/T157721 -- Investigate default JSON
minification behavior in production
28.
https://phabricator.wikimedia.org/T157723 -- ORES swagger is
hard-coded
for wmflabs
29.
https://phabricator.wikimedia.org/T152585 -- rcshow=oresreview is slow
30.
https://phabricator.wikimedia.org/T158862 -- Fix message in
Special:Contributions
31.
https://phabricator.wikimedia.org/T158899 -- Add notice about Dexbot
overwriting manual changes to our tracking table.
32.
https://phabricator.wikimedia.org/T159055 -- Add a notice to
ores-wmflabs-deploy about "experimental" nature
33.
https://phabricator.wikimedia.org/T160192 -- Fix testing issues in
finnish language assets
34.
https://phabricator.wikimedia.org/T160258 -- Fix minor styling issues
with OOJS-UI in wikilabels
Sincerely,
Aaron from the Scoring Platform team
_______________________________________________
Wikitech-l mailing list
Wikitech-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikitech-l
_______________________________________________
Wikimedia-l mailing list, guidelines at:
https://meta.wikimedia.org/
wiki/Mailing_lists/Guidelines and
https://meta.wikimedia.org/
wiki/Wikimedia-l
New messages to: Wikimedia-l(a)lists.wikimedia.org
Unsubscribe:
https://lists.wikimedia.org/mailman/listinfo/wikimedia-l,
<mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe>