not really, since tokenization combines multiple characters