This includes:
- new keyword in CirrusSearch to remove them from search results
- new option in AdvancedSearch to do so
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T304939 Translation pages cluttering various page listings | |||
| Open | Feature | None | T304977 Option to remove translation pages from search results | ||
| Resolved | Reedy | T347429 Store whether a page is a "translation" (sub) page in the database to allow varying SQL queries |
Event Timeline
Comment Actions
CirrusSearch already provides inlanguage:en: https://www.mediawiki.org/wiki/Help:CirrusSearch#Inlanguage
Comment Actions
Reopen.
- Each translatable page also have a /en subpage
- In fact Translate extension allow mark pages in any languages for translation and language of pages can be changed arbitrarily (even without Translate extension)
Comment Actions
Thanks for filing. Just so I better understand your comments above, are you saying that the CirrusSearch language filter is not sufficient because translation pages have English subpages and still show up in search results? I'm trying to better understand why the language filter workaround doesn't work for you.
Comment Actions
Firstly translation pages have English subpages; secondly even uncommon, it is possible for any page in any wiki to be in any languages and there is interface (Special:PageLanguage) to change it.
Comment Actions
Examples:
- m:Tech/News/2022/14/en is found by inlanguage:en (example search), even though it’s a translation (well, kind of).
- m:ItWikiCon/2020 should not be found by inlanguage:en, although it’s not a translation, because its language was manually set to Italian. (Actually it is found, but I’d consider that a bug, not a viable solution for this problem.)
Comment Actions
A big source of frustration here is that for certain keywords on wikis like mediawiki.org or Meta you will get a ton of half-translated Extension:Translate subpages, all containing the English original and this having high specificity scores while actually being utterly irrelevant. Sometimes search is borderline useless because of that.
I think the ideal solution there would be to not index untranslated text on translated subpages, or maybe somehow coalesce translations of the same page into a single search result.