I am trying to dots.ocr image and make a pdf with searchable text using json output but I get only paragraph bbox precision not word precision. Any solutions?