There are a few problems in JavaDoc search related to Unicode (non-ASCII) letters and digits.

 - Unicode (non-ASCII) letters confuse the word boundary algorithm and are findable when they occur in the middle or at the end of a word.
 - Unicode (non-ASCII) letters break the camel-case detection when they immediately follow an upper-case letter in a camel-case signature.
 - (Groups of) digits are not recognized as words by the search algorithm unless they follow a separator or non-letter character.

Some of these issues are implementation bugs, but the JavaDoc search specification should be updated to be more explicit about how these cases are handled. There are also other areas where the search spec could use an update, for instance there are JavaDoc tags other than `{@index}` which create search items and should be mentioned/referred to in the spec.