The recent changes to the query have made things our matching logic a bit too permissive.
From a feedback email:
We looked at searches for:
Mary Barnes (sorted for Newest first), that retrieved over 16,000 records, including everybody called Mary or Marie or Marya;
stereoscopic photographs (sorted for oldest first), which retrieves over 26,000 results including three 15th century manuscripts, none of which appear to have any relevance to the search term;
As a quick fix, we can invert the minimum_should_match percentage from 80% to -20%.
This should mean that instead of requiring 1 token to be present in a 2-token query, or 2 tokens to be present in a 3-token query, etc, we will require all terms to be present. Set at -20%, The minimum_should_match condition will only be applied to queries which are longer than 4 tokens., above which the percentage will be rounded down.
The recent changes to the query have made things our matching logic a bit too permissive.
From a feedback email:
As a quick fix, we can invert the
minimum_should_match
percentage from80%
to-20%
.This should mean that instead of requiring 1 token to be present in a 2-token query, or 2 tokens to be present in a 3-token query, etc, we will require all terms to be present. Set at
-20%
, Theminimum_should_match
condition will only be applied to queries which are longer than 4 tokens., above which the percentage will be rounded down.See the the docs for the minimum_should_match parameter for a more detailed explanation