[Feature request] Index text without diacritics

[Sanqui]

When I search for "zlutoucky", I would like to be able to find "žluťoučký", and vice versa. These results should have a lower relevancy/priority. Would this be possible?

[scambier]

Omnisearch would have to index an additional computed field to hold the converted text without diacritics. Definitely doable, but I'd put that behind a setting, since that would roughly double the indexing duration and increase memory consumption.

I'm not sure how it will work with the in-file search, but I can take a look at it. Anyway, this part needs a rework.

These results should have a lower relevancy/priority

Because the same word with and without diacritics could have 2 different meanings?

[Sanqui]

I don't mind it being behind a setting. It's too bad you may have to convert the text, would be nice if the search library you're using offered diacritics removal alongside stemming possibly lemmatization.

These results should have a lower relevancy/priority

Because the same word with and without diacritics could have 2 different meanings?

Indeed, many such cases. Though I would have to see it in practice to know if it would be usable if the results were ranked the same, I suspect it would be only a minor annoyance.

[scambier]

It's too bad you may have to convert the text, would be nice if the search library you're using offered diacritics removal

Well, I suspect it's specifically because diacritics can change the meaning of a word. I have the same issue in French, but to a lesser degree: the fuzzy matching usually works around it ("creme brulee" will match "crème brulée", but not ""crème brûlée").

I think that a better solution would be a toggle to simply ignore diacritics: normalize all search queries, and notes before indexing.