An Automatic Device for Reading Typographical Texts

摘要

The subject of the paper is a reading machine designed for transferring printed texts into an electronic computer. It can be used in automatic translation systems and information retrieval systems where a comparatively low reading reliability and a speed of about 100 to 200 symbols per second are permissible. The system represents an attempt to apply the methods of machines designed for typescript reading to machines reading printed texts. Some characteristics by which typescript and typographical material differ are presented. The basic aspects of the recognition algorithm are given. A block diagram of the machine is given, together with an explanation of its operation. The memory circuit uses ferrite cores and is divided into four zones; the first of these preserves images of parts of lines; the second remembers operating programs; the third is devoted to symbol standards and their characteristics; the fourth accumulates replies. The section of the machine devoted to the image input uses a travelling beam tube with an independent control system; it transforms the symbol images into a binary code, each work of which is one column of the image and is recorded in an individual cell of the first memory zone. (Author)