Using knowledge in the computer interpretation of handwritten FORTRAN coding sheets

This paper describes the methodology of a project to develop a computer program capable of reading casually hand-printed FORTRAN programs. A major feature of the methodology is the organization of the processes in the recognition mechanism: conventional process organization is rejected as too rigid and inflexible. The proposed process organization, developed by workers in artificial intelligence, depends upon the generation, followed by verification or rejection, of hypotheses about the program being read. An argument is advanced that this will not lead to a combinatorial explosion of search time. The ease with which items of knowledge about FORTRAN and about hand-printing can be incorporated in such a system is stressed. Some details are given about the initial direction of the project work.