I hear that some tools like Huggle have internal storage for when users judge that an ORES score is incorrect. This is valuable data for ORES engineering, and we'd love to help get that data exported into Jade.
Description
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Declined | None | T197098 Use Jade as a repository for existing ORES counterexamples | |||
Resolved | awight | T198207 Write JADE internal APIs to simplify integrations |
Event Timeline
Can you give me some examples how would you interact with this JADE thing? For example, an edit on enwiki is scored by ORES as good quality, but reverted by user in Huggle, meaning it's probably a bad quality edit. What should I do, which API would I call and how the API call format would look?
@Petrb Thanks for the questions! We're still finalizing the schema for judgments, but I'll post the most recent thoughts in T179301: Document JADE judgment structure. It would be great if you wanted to give feedback on the schema before we deploy, of course.
JADE judgments are JSON stored in a new MediaWiki namespace, so it will be edited using the normal content APIs, like action=edit.
Copying from the other task, I've started an explanation here: https://www.mediawiki.org/wiki/JADE/Schema
I found some references to the action=patrol API in Huggle source, but didn't realize until now that it isn't used in practice:
SELECT params["action"], count(*) FROM wmf_raw.ApiAction WHERE year = 2018 AND month = 6 AND day = 20 AND hadError = false AND userAgent regexp '^Huggle' GROUP BY params["action"] LIMIT 1000000 ;
query 196666 watch 37 delete 2 login 79 block 3 rollback 421 compare 39482 edit 425 thank 5
I guess the current storage for "patrolled" really is internal—would you mind pointing to some of the code that does that?
@awight that's because you are probably searching only English wikipedia where patrolling was disabled long time ago, here is very old note from config page:
# patrolling-enabled will make huggle patrol all edits that were marked as good edits # this feature slows huggle down and creates extra traffic when the feature isn't # supported by local wiki installation, so make sure to properly tune this variable according # to wiki config # English wikipedia has patrolling disabled # per https://en.wikipedia.org/w/index.php?title=Wikipedia:Village_pump_%28news%29&diff=9146943&oldid=9146404 # so huggle can't use it
/me crunches on popcorn.
Thanks for the pointer! My query was actually against all wikis, but only for a 24-hour period. It's fine if this integration is disabled everywhere, I'm just curious to understand the current workflows.
If a revision is reviewed in Huggle and is not reverted, I assume you're sharing some type of data with other Huggle users to prevent them from redundant review of this same change?
Huggle has a feature named "HAN", that uses IRC to share instant data between patrollers. Source is at huggle/huggle_ui/vandalnw.cpp.
It looks like clients can't communicate any type of "PATROLLED" message, indicating that a change has been reviewed and is good?
This is sort of on hold, we've decided that we're only interested in Jade-like collaborative judgments, and there's no point in just bulk importing data of potentially lower quality.