Morphological search for non-English languages

• Feb 26, 2015 - 10:13
Type
musescore.org
Severity
S4 - Minor
Status
active
Project

https://twitter.com/almaximal/status/570886725343559680

Some time ago I noticed, that search engine on this site is not so good. It’s not morphological for non-English languages.

> thx for the feedback. Could you give an example of a bad search result? (link)

e.g.:
https://musescore.com/sheetmusic?text=%D0%BB%D1%83%D0%BD%D0%B0
„луна“ is basic form for „луну“ in Russian, so this thing should be found:
https://musescore.com/user/199262/scores/188627

I personally like http://sphinxsearch.com/, but you could use any modern fulltext search engine supporting morphology.


Comments

Ok, we'll investigate if we can improve the morphological analyzers in our search engine. Thanks again for reporting.

If there will be a choice between stemmers and lemmatizers for Russian morphology it’s better to use lemmatizers.
Even though they are heavier while indexing (for memory and processor), results are more precise.