Discovery/Status updates/2017-10-02


This is the weekly update for the week starting 2017-10-02


  • The week of October 9th, we will deploy code to disable the automatic use of UI message fallback languages for linguistic analysis. In most cases, the messaging fallback language is unrelated to the main language of a wiki (e.g., Wolof falls back to French). A fuller description is on MediaWiki,[1] and you can join the discussion there or on the related Phab ticket, T147959.[2]
  • The Analysis team had several meetings to scope out new work to establish a baseline of metrics for the Structured Data on Commons project, you can read more at (task T174519).



  • Trey is investigating linguistic analysis software that could be adapted for use on wiki projects. If you know of any open-source linguistic analysis software for languages that don't already have it on their wiki projects, please drop a note on Phab ticket T171652.[3]
  • David was busy with various tasks:
    • relaxing the default "AND" of the retrieval query filter with minimum_should_match (task T177302)
    • change loading of cross-wiki configuration to use cirrus-dump-config api call (task T156497)
    • finishing up the work to upgrade Elastica to support elastic 5.5.x (will be in production next week (task T174654)
    • and adding Special:Search—add additional namespaces to search results for Persian Wikibooks (task T176908), Hebrew Wikisource (task T176907), and English Wikibooks (task T176906)
  • We also stopped the A/B test that was running on 18 wikis with > 1% of search traffic, analysis will follow (task T177490)


  • Chelsy investigated the full-text search pattern on mobile web and found interesting things: "we think that since people use full-text instead of prefix search results when they have a in-depth information need, and this kind of need is often driven by work/school project, we see more full-text searches on weekdays than weekends on both desktop and mobile web." Read more at (task T176815)


  • Paul kicked off the process to regenerate the map tiles in production (task T176252) that will probably finish early next week. We had to do this manually because of a bug in tileshell/node.js that we're working to resolve (task T175123).