My one thought is that *IF* you define a custom character set definition (which might be part of what is called a 'collation sequence' that made converting a Hebrew character to lower case remove the vowel points (Hebrew doesn't HAVE the concept of Case as far as I know). This might be able to work as the vowel points should be defined as combining characters so they all get bundled up into a single grapheme, which I think is what collation sequences work on. It's possible that this is a common enough request, that there may be a collation sequence already available somewhere that does this.