Halfak added a comment.
https://github.com/wiki-ai/wikiclass/commit/950f693d789f8512e30f483f18e2d13483d13749TASK DETAILhttps://phabricator.wikimedia.org/T157493EMAIL PREFERENCEShttps://phabricator.wikimedia.org/settings/panel/emailpreferences/To: HalfakCc: samuwmde, Lydia_Pintscher, StudiesWorld, A
Glorian_WD added a comment.
@Halfak : could you briefly explain about the bug?TASK DETAILhttps://phabricator.wikimedia.org/T157493EMAIL PREFERENCEShttps://phabricator.wikimedia.org/settings/panel/emailpreferences/To: Glorian_WDCc: samuwmde, Lydia_Pintscher, StudiesWorld, Aklapper, Ricordisamoa, Sum
Halfak added a comment.
There was a bug. So I re-deployed again. See http://labels.wmflabs.org/campaigns/wikidatawiki/53/?campaign=statsTASK DETAILhttps://phabricator.wikimedia.org/T157493EMAIL PREFERENCEShttps://phabricator.wikimedia.org/settings/panel/emailpreferences/To: HalfakCc: samuwmde, Ly
Halfak added a comment.
@Glorian_WD updated the sampling strategy so I re-deployed. See http://labels.wmflabs.org/campaigns/wikidatawiki/52/?campaign=stats
halfak@wikilabels-01:~$ sudo -u www-data /srv/wikilabels/venv/bin/wikilabels new_campaign wikidatawiki "Item quality (5k sample)" item_quali
Halfak added a comment.
halfak@wikilabels-01:~/datasets$ sudo -u www-data /srv/wikilabels/venv/bin/wikilabels new_campaign wikidatawiki "Item quality (5k stratified)" item_quality PrintablePageAsOfRevision 1 10 --config /srv/wikilabels/config/config/
{'active': True, 'labels_per_task': 1, 'tasks_pe
Halfak added a comment.
http://labels.wmflabs.org/campaigns/wikidatawiki/51/?campaign=statsTASK DETAILhttps://phabricator.wikimedia.org/T157493EMAIL PREFERENCEShttps://phabricator.wikimedia.org/settings/panel/emailpreferences/To: HalfakCc: samuwmde, Lydia_Pintscher, StudiesWorld, Aklapper, Ricordis
Halfak added a comment.
I think we'll want a stratified sample for the labeling campaign so that we don't end up with 99% of items in the lowest quality strata. We can probably stratify using some basic heuristics, but which ones? # of statements? We can probably get away without labeling any of