how does full text searching tokenize words ? can it be altered?

I'm getting a handful of 'can not index words longer than 2047 characters' on my `gin` indexes.

1. does this 2047 character count correspond to tokens / indexed words?
2. if so, is there a way to lower this number ?
3. is there a way to profile the index for the frequency of tokens ?


( apologies in advance if this looks familiar, i posted this as part of a larger question last month; everything but
thiswas answered by the list and I can't find answers to this online )