The filter indic_normalization is used in the Hindi and Bengali analysis chains, but the code specifically mentions other scripts, and we should evaluate adding it to languages that use those scripts.

I ended up looking at the code for indic_normalization, and it is definitely pretty complex. It also explicitly calls out several scripts: Bengali, Devanagari, Gujarati, Gurmukhi, Kannada, Malayalam, Odia, Tamil, and Telugu. It would definitely make sense to test whether it is generally useful on languages using those scripts. (It is already used as part of the Bengali and Hindi analyzers. In Bengali it didn't seem to cause any problems, but we didn't necessarily look at a significant amount of text in Gujarati, Kannada, or Tamil scripts when reviewing the Bengali analyzer.)