I'm not sure having statistics with fabrication try to extract text from PDF's would result in any mission-critical reliable data?