OCR: The Kurzweil Data Entry Machine

This article presents an assessment of the Kurzweil Data Entry Machine (KDEM) as an input device for literary and linguistic computing It is based on over five years' experience of providing an optical reading service for academic users The KDEM is an ‘intelligent’ optical character reader in that it is able to ‘learn’ the shapes of the letters in a text It can read proportionately spaced (typeset) text as well as uniformly spaced material (typescript) and can recognise different fonts as well as superscripts, subscripts and underlining Any script in which the letters do not join can be read The accuracy and speed of the reading depend very much on how well the KDEM has been trained Speeds of up to 30 pages per hour can be obtained on a good text with the aid of a built-in lexicon of about 33,000 English words The most suitable types of text are solid blocks or pages in a single typeface which have been printed on relatively smooth paper Dictionaries and bibliographies can be read but are more difficult because of frequent changes of typeface Newspapers and computer printout are unsuitable