We are grateful to the many researchers in OCR who have explored various options in approaching standard problems. First we briefly outline the currently accepted wisdom on the steps necessary for successful document understanding ([]), In subsequent sections we show where our pieces fit in.