BUG: Search Results Showing That Don't Contain The Search Term

[adam-vanbuskirk]

Describe the bug
When searching for a term on our site (topclassactions.com), the results contain many posts that do not contain the search term. You can use kashima as an example. We have a DB of 3GB+ in Prod. The lower life cycle (Dev & Staging) return the correct results. The plugin is not. modified. I cannot figure out why our search returns bad results in Prod.

When this would happen previosuly, I would rebuild the index and it would fix itself. Now, rebuilding or syncing the index doesn't even fix it.

Steps to Reproduce

1. Go to topclassactions.com
2. Click the hour glass at the top right, enter kashima, an press enter.
3. Click on page 6 in the pagination and find the post titled Kellog Hit with Class Action Lawsuit Over 'Real Fruit' Claims and do a control F for kashima and you won't find it. There are several other examples in that same search feed that don't contain the term.

Expected behavior
It should only return posts that contain the search term, in this case kashima

Screenshots

Environment information

• WordPress version: 5.9.3
• ElasticPress version: 4.2.2
• Elasticsearch version: 7.9.3
• What ElasticPress features do you have currently active? Post Search, Facets

[felipeelia]

Hi @adam-vanbuskirk,

This is not a bug. By default, Elasticsearch will try to find similar words and not only exact matches. Looking at your screenshot I can see Kashi and that is just two letters different from Kashima.

If you want to disable that behavior, you can use the following snippet:

add_filter( 'ep_match_fuzziness', '__return_zero' );

To read more about Elasticsearch Fuzziness, you can access their docs.

[kevinlisota]

@felipeelia the ep_match_fuzziness filter is deprecated, replaced by ep_post_match_fuzziness. However, returning zero to the new filter does not turn off fuzzy matching. What is the equivalent snippet for 4.3.0?