If you prefix search for "gold" on Wikidata in English, you get Q897. However, if you switch to Russian and search for "gold", you still get matches, but Q897 is nowhere to be seen. This looks wrong - Q897 should still show higher on account of it's scoring by links, etc.
See: https://www.wikidata.org/wiki/Wikidata:Project_chat#Search_misses_top_item
Description
Description
Details
Details
Related Changes in Gerrit:
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| [search] better tuning with language fallbacks for prefix search | mediawiki/extensions/Wikibase | master | +84 -84 |
Event Timeline
Comment Actions
We really need a centralized place to store all these queries and expected results with different parameters. The key to making effective search is to have a set of queries and known good results, and then be able to evaluate changes to the system in how it affects all of those queries.
Comment Actions
Looking at http://wikidata-wdsearch.wmflabs.org/w/api.php?useCirrus=1&action=wbsearchentities&search=gold&format=json&limit=10&type=item&language=ru&uselang=en&cirrusDumpResult=1&cirrusExplain=pretty - searching for "gold" in Russian - looks like top items are getting high scores on language match mainly.
Comment Actions
@debt i think we should plan some work on search tuning for Wikidata on Q3, yes. @Lydia_Pintscher has also mentioned there were some complaints and we're planning to assemble them and do another round of tuning.
Comment Actions
Change 397887 had a related patch set uploaded (by DCausse; owner: DCausse):
[mediawiki/extensions/Wikibase@master] [search] better tuning with language fallbacks for prefix search
Comment Actions
Change 397887 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] [search] better tuning with language fallbacks for prefix search