TODO: There are still errors in parsing the round marks, as some are not offset correctly by pdftotext.