Here matches on multipart IDs (shelfmark & collection path) are boosted less than other ID matches because the nature of these identifiers is that small parts of them may collide with valid searches (for example "of", "wa", "sa") pushing matching results for this ID parts above other results in a way that may confuse users.
What does this change?
This change updates search queries to include the shelfmark value added by https://github.com/wellcomecollection/catalogue-pipeline/pull/2693.
We add a new block to the works query:
Here matches on multipart IDs (shelfmark & collection path) are boosted less than other ID matches because the nature of these identifiers is that small parts of them may collide with valid searches (for example "of", "wa", "sa") pushing matching results for this ID parts above other results in a way that may confuse users.
This change was revealed by, and tested with
wellcomecollection/rank
: https://github.com/wellcomecollection/rank/pull/111/filesPart of https://github.com/wellcomecollection/catalogue-pipeline/issues/2453
Checklist
How to test
How can we measure success?
Collection staff are more easily able to find the works they are looking for.
Have we considered potential risks?
We should have rank tests to ensure this does not change search query results in unexpected and detrimentatl ways before merging to mititigate risk.