Meilisearch return results with a completely different word in it

[sagar2911]

Describe the bug
See screenshots, When I look up the term 'estates at weddington', the first result is correct but it also has 41 more results that don't really match like Estates at Rockhill and Estates at Hutsfield

To Reproduce
Steps to reproduce the behavior (not sure as we have our own data but):

1. Create some documents with similar substrings.
2. search for one of the strings and it should return some other ones that are similar but not the same.

Expected behavior
Only have results with at most 2 letters of typo like the documentation says.

Screenshots

I wondered if maybe the word weddington is nested deep within the other documents but that's not the case

Meilisearch version: [e.g. v0.25.2]

Thank you

[ManyTheFish]

Hey @sagar2911, this is not a bug, but the behavior of Meilisearch.

Meilisearch will always try to return the number of documents you requested, sorted by the ranking rules:
https://docs.meilisearch.com/learn/core_concepts/relevancy.html#ranking-rules

That means that if you request 20 documents, and only 1 document match perfectly (all words, no typo...);
Meilisearch will try to allow more and more typos, then it will try to remove words in the query...

For example, in your screenshots:

• screenshot 1: Meilisearch is searching the 3 words then 2 then 1
• screenshot 2: because you provided only one word, Meilisearch will stop when 2 typos are allowed.

I hope this answers your issue. 🙂

PS for @gmourier: This issue "shows" unintuitive behavior of the ranking rule Words, we may enhance it and allow our users to deactivate optional words

[gmourier]

Hi @sagar2911 and @ManyTheFish!

Correct me if i'm wrong but an optionalWords option at false would strictly match documents having all the queried words with 0 typo?

Does this option deactivate the typo tolerance feature in this case?

Thanks

[ManyTheFish]

@gmourier, no, it would avoid removing words in the query.
This would be the only way to deactivate this behavior, removing Words ranking rules wouldn't do it.

[MarinPostma]

@ManyTheFish, why couldn't remove the Word criterion do that? It seems that this is the intuitive thing to do, and this seems validated by user behavior. What is blocking for this issue to make progress?

[ManyTheFish]

@MarinPostma yes, I agree, I was just describing the current behavior. 😊

[curquiza]

Closing this issue since an answer has been given. I linked this issue to the related discussion on the product side meilisearch/product#255 for traceability

[curquiza]

Hello @sagar2911 and everyone here! 👋

We are working on introducing the relevancy customization during the search with Meilisearch, and we just released a first prototype that allows you, for example, only to return documents containing ALL the query words! 🔥

It would be super helpful if you could let us know if this helps with your issue.
To know more about the prototype and how to use it, take a look at the dedicated discussion.

[miagg]

OMG that would be awesome! I've been looking for this for a while. I use Meilisearch from a backend to match product titles from parsed documents where I expect ONLY a single hit as a result and I often get many hits which is useless for my case since the first hit is not necessary the right one. This is because of large differentiation in product titles among retailers. This prototype would help a lot in my case. I do hope to see in the future some kind of a relevancy threshold where Meilisearch would not return results below the given value.

[curquiza]

Hello @sagar2911, @miagg and everyone here 👋

We have just released v0.29.0rc1, which is a release candidate of v0.29.0 🔥

It introduces a new search parameter named matchingStrategy which allows you to change the default Meilisearch behavior about retrieving documents.

For now, it accepts 2 strategies:

• last: the default strategy. If no document contains all the given terms, the engine removes the last query term and so on.
• all: All the query terms must be present in a document to make it a search result.

How to use it:

curl \
  -X POST 'http://localhost:7700/indexes/movies/search' \
  -H 'Content-Type: application/json' \
  --data-binary '{ "q": "american ninja", "matchingStrategy": "all" }'

Binaries are attached to the release, or you can use the docker image:

docker run -it --rm \
    -p 7700:7700 \
    getmeili/meilisearch:v0.29.0rc1

Let us know about any bugs or feedback! 😄 It would be really helpful.

FYI, the official v0.29.0 release will be available on 3rd October.