Search Platform/Weekly Updates/2025-06-06
Highlights
Shipped: did we release anything this week?
- We now have a pretty cool Sankey diagram on how users flow through search (there is also a more complete / complex version). Diagrams are cool! Sankey diagrams doubly so! This diagram helps tell a more complete story than what we see on our Search Metrics dashboard. For example, the dashboard shows a ~50% abandonment on full text search, which seems pretty bad. The full story is that users usually first go through auto complete search, which has a very high success rate (roughly 90%). Only the queries that are unsuccessful at that point will end up on full text search and the Search Engine Result Page (SERP). So full text search mostly sees the "hard" queries. Those sankey diagrams also make it quite clear that advanced users (users with 1000+ edits) are a lot more likely to use full text, and probably the advanced features of search.
- Analysis of Indic Normalization is complete. 27 languages were analyzed, changes to the analysis chain was done for 13 of them. This improve search recall for those languages. Read the full analysis!
Blockers: does this essential workstream have any unresolved blockers or dependencies? is anything preventing us from doing our work?
N/A
Lessons learned: Did we learn anything in the course of doing this work that can be applied to other work?
N/A
Community collab: Did we do anything in this essential workstream this week in collaboration with the community or because the community asked us to?
N/A
What else was accomplished in this essential workstream this week?
Sankey diagram
Misc / Operations
- T395677 Search backend error: illegal_argument_exception: field value function must not produce negative scores (Resolved)
- T391792 Align search platform DAGs to DPE best practices (Resolved)
- T392525 Investigate creation of a Sankey diagram for user interaction with search (Resolved)
- T391738 outreach.wikimedia.org is not allowlisted for an mwapi Wikidata query (Resolved)