[Wikidata-bugs] [Maniphest] T349246: Bad ranking of Wikidata item search results on Special:Search when non-default namespaces are included

[Gehel]

Gehel closed this task as "Declined".
Gehel added a comment.


  This is working as designed. A complete rework of how we merge different 
search results is out of scope.

TASK DETAIL
  https://phabricator.wikimedia.org/T349246

EMAIL PREFERENCES
  https://phabricator.wikimedia.org/settings/panel/emailpreferences/

To: Gehel
Cc: Gehel, TJones, Aklapper, Nikki, Danny_Benjafield_WMDE, Astuthiodit_1, 
karapayneWMDE, Invadibot, maantietaja, ItamarWMDE, Akuckartz, Nandana, Lahi, 
Gq86, GoranSMilovanovic, QZanden, EBjune, LawExplorer, _jensen, rosalieper, 
Scott_WUaS, Wikidata-bugs, aude, Mbch331
___
Wikidata-bugs mailing list -- wikidata-bugs@lists.wikimedia.org
To unsubscribe send an email to wikidata-bugs-le...@lists.wikimedia.org

[Wikidata-bugs] [Maniphest] T349246: Bad ranking of Wikidata item search results on Special:Search when non-default namespaces are included

[TJones]

TJones added a comment.


  Somewhat unfortunately, this is the expected behavior.
  
  There's a specialized query for Wikidata items that doesn't work with other 
namespaces. When you include other namespaces, we have to fall back to a less 
good query that works consistently across namespaces and allows us to merge 
results from multiple namespaces. That query also allows us to search all 
namespaces with a single request. In the //Finno-Ugric// case, there aren't any 
results from other namespaces, but you still get the ranking from the less good 
query.
  
  There are approaches to merging result lists that were scored with different 
scoring methods, but we haven't seriously investigated doing that for on-wiki 
search. There is also the expense of running multiple queries before you can 
merge their results, which can make such queries much more expensive.
  
  Reframing the situation (and this is approximately what actually happened 
from the programmers' point of view), if you limit yourself to default 
namespaces, we can run a specialized query that uses much better ranking for 
Wikidata items.

TASK DETAIL
  https://phabricator.wikimedia.org/T349246

EMAIL PREFERENCES
  https://phabricator.wikimedia.org/settings/panel/emailpreferences/

To: TJones
Cc: TJones, Aklapper, Nikki, Danny_Benjafield_WMDE, Astuthiodit_1, 
karapayneWMDE, Invadibot, maantietaja, ItamarWMDE, Akuckartz, Nandana, Lahi, 
Gq86, GoranSMilovanovic, QZanden, EBjune, LawExplorer, _jensen, rosalieper, 
Scott_WUaS, Wikidata-bugs, aude, Mbch331
___
Wikidata-bugs mailing list -- wikidata-bugs@lists.wikimedia.org
To unsubscribe send an email to wikidata-bugs-le...@lists.wikimedia.org

[Wikidata-bugs] [Maniphest] T349246: Bad ranking of Wikidata item search results on Special:Search when non-default namespaces are included

[Nikki]

Nikki created this task.
Nikki added projects: Wikidata, Discovery-Search.
Restricted Application added a subscriber: Aklapper.

TASK DESCRIPTION
  When searching Wikidata using Special:Search, the ranking of the results is 
bad when any non-default namespaces are included.
  
  For example, when searching for "Finno-Ugric", the expected top result would 
be Q79890  which is the only item with an 
English label that matches exactly.
  
  Searching in a private window:
  
  - Main and property namespaces (the default) 
:
 1st result
  - Only the main namespace 
:
 1st result
  - Main, property and talk namespaces 
:
 38th result
  - Main, property and lexeme namespaces 
:
 38th result
  
  In all four cases, the first two pages of results are all items, but the 
order is completely different. This is unintuitive - searching more namespaces 
should not affect how well search terms match within a namespace.

TASK DETAIL
  https://phabricator.wikimedia.org/T349246

EMAIL PREFERENCES
  https://phabricator.wikimedia.org/settings/panel/emailpreferences/

To: Nikki
Cc: Aklapper, Nikki, Danny_Benjafield_WMDE, Astuthiodit_1, karapayneWMDE, 
Invadibot, maantietaja, ItamarWMDE, Akuckartz, Nandana, Lahi, Gq86, 
GoranSMilovanovic, QZanden, EBjune, LawExplorer, _jensen, rosalieper, 
Scott_WUaS, Wikidata-bugs, aude, Mbch331
___
Wikidata-bugs mailing list -- wikidata-bugs@lists.wikimedia.org
To unsubscribe send an email to wikidata-bugs-le...@lists.wikimedia.org