[Wikisource-l] Changes coming to ORES model for prediction of page levels in French Wikisource

[Amir Sarabadani]

Hey,
ORES has a model that predicts level of a document for French Wikisource.
For example it predicts this version of "Le Comte de Monte-Cristo" [1]
needs more proofreading [2]. [2] is an example response which you can see
predication classes are either "0", "1", "3" or "4" (with this mapping
[3]). This is suboptimal, counter-intuitive and the most importantly, it's
causing EventStream to fail to emit new scores for French Wikisource [4].
So we are changing it the "0" class to "without_text", "1" to
"not_proofread" and so on in the next week.

If you are using this class, be alarmed about this change and please align
your tools. We are very sorry for any inconvenience.

[1]:
https://fr.wikisource.org/w/index.php?title=Le_Comte_de_Monte-Cristo/Chapitre_33&oldid=5671101
[2]: https://ores.wikimedia.org/v3/scores/frwikisource/5671101
[3]: https://phabricator.wikimedia.org/T195979#4249618
[4]: https://phabricator.wikimedia.org/T195979

Best
--
Amir Sarabadani
Software Engineer

Wikimedia Deutschland e. V. | Tempelhofer Ufer 23-24 | 10963 Berlin
Tel. [phone]
http://wikimedia.de

Stellen Sie sich eine Welt vor, in der jeder Mensch an der Menge allen
Wissens frei teilhaben kann. Helfen Sie uns dabei!
http://spenden.wikimedia.de/

Wikimedia Deutschland – Gesellschaft zur Förderung Freien Wissens e. V.
Eingetragen im Vereinsregister des Amtsgerichts Berlin-Charlottenburg unter
der Nummer 23855 B. Als gemeinnÃŒtzig anerkannt durch das Finanzamt fÃŒr
Körperschaften I Berlin, Steuernummer 27/029/42207.