User Story: As a user of Bengali-language wikis, I want to have better Bengali language analysis so I see better search results (particularly, better recall).
Elasticsearch provides a Bengali language analyzer, but we don't currently use it for Bengali-language projects. We should enable it, have the performance verified by speakers, and then unpack it.
- Bengali speakers verify reasonable performance of the stemmer
- Unpacked analyzer performs the same as the monolithic version (without general upgrades).
- Upgraded analyzer either has no unexpected impact (we know what to expect from ICU norm and homoglyph norm, for example), or the impact is reviewed by a speaker of the language.
- Analysis changes are deployed, a re-indexing sub-task is created off this task's parent (T272606), and linked to in T147505.