Can I turn off Korean tokenizer?

[datamaker]

We are stuck at version v0.29.
The search quality is not good because of the phenomenon that Hangul morpheme analysis is too finely divided.
However, I want to use all the features of v1.3, but I can't.

We believe that the search quality is not as good as it is currently based on whitespace.
#623
#564

Will it be accepted if I PR the ability to turn the tokenizer on and off with an environment variable?
meilisearch/meilisearch#3905

[ManyTheFish]

Hello @datamaker, it's unfortunate that the tokenization of Hangul doesn't suit your need, we want to have good support in the Korean Language, could you please do several tests on your side:

1. change the order of your ranking rules to ["word", "proximity", "exactness", "typo", "attribute"]
2. recompile Meilisearch with the following command line:
   a) cargo build --release --no-default-features --features "analytics mini-dashboard"
   b) /target/release/meilisearch
   c) reindex your dataset

Let me know if it enhances your use case, and which one is the most efficient. Don't hesitate to give examples.

Sorry for this inconvenience.