Wikimedia Search Platform
The Search Platform team (part of Wikimedia Technology) is responsible for maintaining and enhancing the various Search features and APIs for MediaWiki. This includes the CirrusSearch extension which relies on Elasticsearch, the search backend used at the Wikimedia Foundation to support Wikimedia projects and Wikidata Query Service, the SPARQL endpoint used to query Wikidata.
Our mission is to help people easily discover knowledge on Wikipedia and its sister projects by providing tools and infrastructure for casual readers and expert users with precise needs, while maintaining a strong emphasis on privacy.
- We operate and maintain a disparate collection of production services related to content discovery, enabling the wiki community to find information that is not available through simply following links. We also provide a platform on which other people can create tools to support editing and other workflows.
- We provide an open-source search engine, backed by an inverted index for non-structured on-wiki data. We work to develop more sophisticated searching with machine learning and natural language processing.
- We provide a SPARQL-based query service for Wikidata, encouraging users to capitalize on this vast store of computer-readable structured data for use on-wiki and in knowledge discovery.
- We endeavor to support underserved wiki communities, and we rely on those communities to help us understand their needs and evaluate potential solutions, especially with respect to underserved languages.
- We prioritize privacy for logged-in users and anonymity for logged-out users over almost everything else, even when it slows down or complicates development or hinders our ability to collect or use data.
The Search Platform team's goals are part of the entire Technology Department's goals. You can links to the current quarterly goals here. (Note that Q1 is July–September for random historical reasons.)
Wikidata Query Service (WDQS)
The Wikidata query service allows for searching structured data on Wikidata. It also provides an API through which tools can access Wikidata. Our current work is tracked on the Discovery-Search workboard (see also our WDQS backlog board) and weekly deployments of WDQS are documented on wikitech:Deployments. A public WDQS Analytics Dashboard is used to monitor and analyze the impact of our efforts.
Application Programming Interfaces (APIs) provide developers ways to interact with the MediaWiki software.
API:Search and discovery lists the search APIs available and in development.
This list was last updated on December 3, 2020.
- Erik Bernhardson, Tech Lead, Staff Software Engineer
- David Causse, Senior Software Engineer
- Trey Jones, Senior Computational Linguist
- Ryan Kemper, Site Reliability Engineer
- Zbyszko Papierski, Senior Software Engineer
- Maryum Styles, Software Engineer
- Guillaume Lederrey, Engineering Manager
- Carly Bogen, Program Manager
- Mike Pham, Senior Product Manager
Search Platform - A public mailing list about the Wikimedia Search Platform team and projects (formerly Discovery Department). Examples of topics would include:
- Announcements, including major upcoming initiatives, completed major releases, quarterly or annual plans, requests for feedback or input
- Technical discussions and brainstorming regarding our work:
- Search, Elastic, Cirrus, the Relevance Forge, and other relevant subjects
- Our dashboards or related analysis
- Other team news, such as changes to team structure, significant changes to processes, changes in how we use phabricator or other tools like gerrit
The Search Platform Team usually holds office hours on the first Wednesday of each month. Come talk to us about anything related to Wikimedia search! Feel free to add your items to the Etherpad Agenda for the next meeting.
Weekly status updates
- San Francisco
- Directly relevant
- Indirectly related (these sorts of meetup groups attract smart/enthusiastic people who like to spend their free time learning and solving problems)
The Search Platform team uses a Scrumban process, which is a hybrid of Scrum and Kanban. It is described here: Search Platform Process.
Conferences, gatherings, and other events
- none scheduled due to COVID-19 global pandemic
- All Hands - January 2018
- Hackathon 2018 - 18 – 20 May 2018
- Wikimania 2018 - July 18-22, 2018
- 17th International Semantic Web Conference (ISWC 2018) - October 8-12, 2018
- October 22 - 25, 2018, Wikimedia Technical Conference (WMTechConf, formerly known as DevSummit) in Portland, Oregon
- Late January / early February 2019, All-Hands, San Francisco
- May 2019, Hackathon, Prague
- Late January / early February 2020, All-Hands, San Francisco
- May 2020, Hackathon, Tirana, Albania (virtual due to COVID-19 global pandemic)
Docs and Other Links
- Help:CirrusSearch - Information on how the Wikimedia search works.
- Completion Suggester - incremental search
- Search Glossary—a place for definitions, context, and links for terms we use that other people may not be familiar with
- Testing Search—testing search changes is complicated!
- Elasticsearch stats on Grafana
- Data access and analysis guidelines used by the Search Platform team around data sources, or by other teams around Search Platform data sources, are documented on Meta
- BrowserBot - a browser test bot for search
- Top Unsuccessful Search Queries - The difficulties in creating a list of unsuccessful search results
- Cross-wiki Search Result Improvements
- TextCat—a software component used for doing language detection
- Data Analysis Archive
The Search Platform team was formerly part of the Discovery Department in Audiences; but, as part of the re-organization (tune-up) of June 2017, the Search Platform team is now part of Technology. Pages of historical note:
Useful reference for who can deploy code. It's nice to know whom to bug if you need something:
The Search Platform team supports the following code:
|Wikidata Query Service||https://phabricator.wikimedia.org/diffusion/WDQR/|
|Wikidata Query Service GUI||https://phabricator.wikimedia.org/diffusion/WDQG/|
|WDQS GUI deployment|
|Lucene Explain Parser||https://phabricator.wikimedia.org/diffusion/WLEP/|